Odyssey: Cross-Domain Technical Systems

• Odyssey is a cross-domain label that describes both tangible tools and metaphorical journeys across various research fields.
• It encompasses concrete systems such as datasets, code frameworks, and experimental challenges in speech, robotics, astrophysics, and more.
• The term also symbolizes staged methodological explorations, reflecting multi-module workflows and iterative design in complex technical landscapes.

to=arxiv_search ＿影音先锋 彩神争霸 彩神争霸大发json {"query":"Odyssey arXiv Odyssey 2024 challenge Odyssey tool code dataset framework", "max_results": 10, "sort_by": "relevance"}򐚨json պատասխան to=arxiv_search 彩神争霸如何json [{"arxiv_id":"(Costa et al., 2024)","title":"Double Multi-Head Attention Multimodal System for Odyssey 2024 Speech Emotion Recognition Challenge","authors":["S. Park","M. Al Ismail","P. Lopez-Otero","L. Docio-Fernandez","C. Garcia-Mateo"],"abstract":"As computer-based applications are becoming more integrated into our daily lives, the importance of Speech Emotion Recognition (SER) has increased significantly. Promoting research with innovative approaches in SER, the Odyssey 2024 Speech Emotion Recognition Challenge was organized as part of the Odyssey 2024 Speaker and Language Recognition Workshop. In this paper we describe the Double Multi-Head Attention Multimodal System developed for this challenge. Pre-trained self-supervised models were used to extract informative acoustic and text features. An early fusion strategy was adopted, where a Multi-Head Attention layer transforms these mixed features into complementary contextualized representations. A second attention mechanism is then applied to pool these representations into an utterance-level vector. Our proposed system achieved the third position in the categorical task ranking with a 34.41% Macro-F1 score, where 31 teams participated in total.","categories":["eess.AS","cs.CV","cs.LG"],"published":"2024-06-15"},{"arxiv_id":"(Misback et al., 2023)","title":"Odyssey: An Interactive Workbench for Expert-Driven Floating-Point Expression Rewriting","authors":["P. Panchekha","A. Sanchez-Stern","J. Qiu","Z. Tatlock"],"abstract":"In recent years, researchers have proposed a number of automated tools to identify and improve floating-point rounding error in mathematical expressions. However, users struggle to effectively apply these tools. In this paper, we work with novices, experts, and tool developers to investigate user needs during the expression rewriting process. We find that users follow an iterative design process. They want to compare expressions on multiple input ranges, integrate and guide various rewriting tools and understand where errors come from. We organize this investigation's results into a three-stage workflow and implement that workflow in a new, extensible workbench dubbed Odyssey. Odyssey enables users to: (1) diagnose problems in an expression, (2) generate solutions automatically or by hand, and (3) tune their results. Odyssey tracks a working set of expressions and turns a state-of-the-art automated tool \"inside out,\" giving the user access to internal heuristics, algorithms, and functionality. In a user study, Odyssey enabled five expert numerical analysts to solve challenging rewriting problems where state-of-the-art automated tools fail. In particular, the experts unanimously praised Odyssey's novel support for interactive range modification and local error visualization.","categories":["cs.HC","cs.PL"],"published":"2023-05-17"},{"arxiv_id":"(Pu et al., 2016)","title":"Odyssey: A Public GPU-Based Code for General-Relativistic Radiative Transfer in Kerr Spacetime","authors":["Y. Pu","Y. Mizuno","Z. Younsi","K. Yaqoob","S. Noble","H. Wu","K. Fuerst","K. Takahashi","K. Makishima"],"abstract":"General-relativistic radiative transfer (GRRT) calculations coupled with the calculation of geodesics in the Kerr spacetime are an essential tool for determining the images, spectra and light curves from matter in the vicinity of black holes. Such studies are especially important for ongoing and upcoming millimeter/submillimeter (mm/sub-mm) Very Long Baseline Interferometry (VLBI) observations of the supermassive black holes at the centres of Sgr A^* and M87. To this end we introduce Odyssey, a Graphics Processing Unit(GPU)-based code for ray tracing and radiative transfer in the Kerr spacetime. On a single GPU, the performance of Odyssey can exceed 1 nanosecond per photon, per Runge-Kutta integration step. Odyssey is publicly available, fast, accurate, and flexible enough to be modified to suit the specific needs of new users. Along with a Graphical User Interface (GUI) powered by a video-accelerated display architecture, we also present an educational software tool, Odyssey_Edu, for showing in real time how null geodesics around a Kerr black hole vary as a function of black hole spin and angle of incidence onto the black hole.","categories":["astro-ph.HE","astro-ph.IM","physics.comp-ph"],"published":"2016-01-09"},{"arxiv_id":"(Chatzakis et al., 2023)","title":"Odyssey: A Journey in the Land of Distributed Data Series Similarity Search","authors":["K. Zoumpatianos","S. Idreos","T. Palpanas"],"abstract":"This paper presents Odyssey, a novel distributed data-series processing framework that efficiently addresses the critical challenges of exhibiting good speedup and ensuring high scalability in data series processing by taking advantage of the full computational capacity of modern clusters comprised of multi-core servers. Odyssey addresses a number of challenges in designing efficient and highly scalable distributed data series index, including efficient scheduling, and load-balancing without paying the prohibitive cost of moving data around. It also supports a flexible partial replication scheme, which enables Odyssey to navigate through a fundamental trade-off between data scalability and good performance during query answering. Through a wide range of configurations and using several real and synthetic datasets, our experimental analysis demonstrates that Odyssey achieves its challenging goals.","categories":["cs.DC","cs.DB"],"published":"2023-01-26"},{"arxiv_id":"(Zhou et al., 29 Aug 2025)","title":"Odyssey: Adaptive Policy Selection for Resilient Distributed Training","authors":["C. Wu","Y. Shen","Z. Wang","S. Yi","W. Zhou"],"abstract":"Training LLMs faces frequent interruptions due to various faults, demanding robust fault-tolerance. Existing backup-free methods, such as redundant computation, dynamic parallelism, and data rerouting, each incur performance penalties, whether from ongoing overhead, lengthy reconfigurations, or post-recovery inefficiencies. We propose Odyssey, an adaptive fault-tolerant system that intelligently selects optimal recovery strategies when a failure occurs. Odyssey achieves this through a unified performance model, expedient execution plan search, accurate performance estimation, and efficient communication optimizations. Experiments on a 32-card cluster show that Odyssey maintains a performance gap of within 11.00% between post-recovery and failure-free training, while preserving model convergence and efficient memory usage. Compared to state-of-the-art methods, Odyssey achieves up to 1.229x and 1.355x higher average throughput than Oobleck and Recycle, respectively.","categories":["cs.DC","cs.LG"],"published":"2025-08-29"},{"arxiv_id":"(Wang et al., 11 Aug 2025)","title":"ODYSSEY: Open-World Quadrupeds Exploration and Manipulation for Long-Horizon Tasks","authors":["K. Wang","Z. Han","Y. Zhou","Z. Li","Q. Zhou","H. Xiang","X. Xie","Y. Liu","H. Xiao","Y. Wang","H. Zhang","Y. Liao","Y. Wang","H. Dong","Y. Wang"],"abstract":"Language-guided long-horizon mobile manipulation has long been a grand challenge in embodied semantic reasoning, generalizable manipulation, and adaptive locomotion. Three fundamental limitations hinder progress: First, although LLMs have improved spatial reasoning and task planning through semantic priors, existing implementations remain confined to tabletop scenarios, failing to address the constrained perception and limited actuation ranges of mobile platforms. Second, current manipulation strategies exhibit insufficient generalization when confronted with the diverse object configurations encountered in open-world environments. Third, while crucial for practical deployment, the dual requirement of maintaining high platform maneuverability alongside precise end-effector control in unstructured settings remains understudied. In this work, we present ODYSSEY, a unified mobile manipulation framework for agile quadruped robots equipped with manipulators, which seamlessly integrates high-level task planning with low-level whole-body control. To address the challenge of egocentric perception in language-conditioned tasks, we introduce a hierarchical planner powered by a vision-LLM, enabling long-horizon instruction decomposition and precise action execution. At the control level, our novel whole-body policy achieves robust coordination across challenging terrains. We further present the first benchmark for long-horizon mobile manipulation, evaluating diverse indoor and outdoor scenarios. Through successful sim-to-real transfer, we demonstrate the system's generalization and robustness in real-world deployments, underscoring the practicality of legged manipulators in unstructured environments. Our work advances the feasibility of generalized robotic assistants capable of complex, dynamic tasks.","categories":["cs.RO","cs.AI"],"published":"2025-08-11"},{"arxiv_id":"(Mahadevan, 25 Jun 2026)","title":"Odyssey: Constructing Verifiable Local Truth-Preserving Foundation Models","authors":["A. Q. V. Luu","S. Sosnowski","Y. Zhang","N. Ruchansky","L. Demeo","P. Frazier","J. Haddad","Z. Ju","S. Nelson","L. Nie","P. Ozyegin","A. Parikh","A. Puli","C. Riedel","S. Santhanam","S. Selvin","J. Shanmugam","W. Shen","R. Vyas","R. S. Yu","B. Zhang","C. C. Aggarwal","S. Venkatasubramanian","S. W. Sutherland"],"abstract":"We introduce a categorical framework called ODYSSEY for constructing verifiable, local truth-preserving foundation models as compositions of foundries: building-block architectural components that specify a cover of local contexts, local representation families, restriction maps, gluing rules, obstruction policies, update obligations, and human-facing views. A foundry is an organized sheaf of knowledge that carries within it an argumentation component. Concrete foundries are built from generic foundries such as evidence/argument, operational decision, institutional/financial, market meaning, scientific challenge, research-program, assistant-build, and evaluation-harness foundries. Universal Foundry Learning (UFL) formalizes foundry construction as a composition of left and right Kan extensions, with left Kan extension rolling local artifacts into candidate foundries and right Kan extension enforcing the restriction, gluing, obstruction, and argumentation conditions required for promotion. Foundry SQL (FSQL) is a small typed query surface for slicing maintained foundry artifacts that uses TICKET (Topos Integration using Causal Kan Extension Transformers) certification for admitting external or pre-built models into durable ODYSSEY state. ODYSSEY is fully implemented and tested across a wide spectrum of concrete foundries, showing that the same categorical machinery supports domain construction, artifact replay, sheaf diagnostics, grounded Toulmin/local-LLM scrutiny, residual-obstruction ledgers, and optimized TICKET-compatible causal-claim extraction across heterogeneous sources. This paper is to be presented as a 2.5 hour tutorial at ICML 2026. The tutorial home page is at https://bit.ly/4ajS0nA.","categories":["cs.AI","cs.LG"],"published":"2026-06-25"},{"arxiv_id":"([2512.14428](/papers/2512.14428))","title":"Odyssey: An Automotive Lidar-Inertial Odometry Dataset for GNSS-denied situations","authors":["F. Y. R. Klaus","G. Schierle","J. Schwenk","M. Magnusson","J. Behley","C. Stachniss"],"abstract":"The development and evaluation of Lidar-Inertial Odometry (LIO) and Simultaneous Localization and Mapping (SLAM) systems requires a precise ground truth. The Global Navigation Satellite System (GNSS) is often used as a foundation for this, but its signals can be unreliable in obstructed environments due to multi-path effects or loss-of-signal. While existing datasets compensate for the sporadic loss of GNSS signals by incorporating Inertial Measurement Unit (IMU) measurements, the commonly used Micro-Electro-Mechanical Systems (MEMS) or Fiber Optic Gyroscope (FOG)-based systems do not permit the prolonged study of GNSS-denied environments. To close this gap, we present Odyssey, a LIO dataset with a focus on GNSS-denied environments such as tunnels and parking garages as well as other underrepresented, yet ubiquitous situations such as stop-and-go-traffic, bumpy roads and wide open fields. Our ground truth is derived from a navigation-grade Inertial Navigation System (INS) equipped with a Ring Laser Gyroscope (RLG), offering exceptional bias stability characteristics compared to IMUs used in existing datasets and enabling the prolonged and accurate study of GNSS-denied environments. This makes Odyssey the first publicly available dataset featuring a RLG-based INS. Besides providing data for LIO, we also support other tasks, such as place recognition, through the threefold repetition of all trajectories as well as the integration of external mapping data by providing precise geodetic coordinates. All data, dataloader and other material is available online at https://odyssey.uni-goettingen.de/ .","categories":["cs.RO","cs.CV"],"published":"2025-12-16"},{"arxiv_id":"(Kim et al., 2024)","title":"A Data-Driven Odyssey in Solar Vehicles","authors":["J. Hwang","C. Kim","J. Kim"],"abstract":"Solar vehicles, which simultaneously produce and consume energy, require meticulous energy management. However, potential users often feel uncertain about their operation compared to conventional vehicles. This study presents a simulator designed to help users understand long-distance travel in solar vehicles and recognize the importance of proper energy management. By utilizing Google Maps data and weather information, the simulator replicates real-world driving conditions and provides a dashboard displaying vehicle status, updated hourly based on user-inputted speed. Users can explore various speed policy scenarios and receive recommendations for optimal driving strategies. The simulator's effectiveness was validated using the route of the World Solar Challenge (WSC). This research enables users to monitor energy dynamics before a journey, enhancing their understanding of energy management and informing appropriate speed decisions.","categories":["eess.SY","cs.CY"],"published":"2024-10-23"},{"arxiv_id":"(Ma, 2020)","title":"Segmentation Loss Odyssey","authors":["J. Ma","Y. He","F. Li","L. Han","C. You","B. Wang","Y. Gao"],"abstract":"Loss functions are one of the crucial ingredients in deep learning-based medical image segmentation methods. Many loss functions have been proposed in existing literature, but are studied separately or only investigated with few other losses. In this paper, we present a systematic taxonomy to sort existing loss functions into four meaningful categories. This helps to reveal links and fundamental similarities between them. Moreover, we explore the relationship between the traditional region-based and the more recent boundary-based loss functions. The PyTorch implementations of these loss functions are publicly available at https://github.com/JunMa11/SegLoss.","categories":["cs.CV","eess.IV"],"published":"2020-05-27"}] Odyssey is a recurrent designation in arXiv literature rather than a single canonical dataset, model, or software stack. In current usage it can denote a workshop venue and challenge framework, a public scientific code, an interactive numerical workbench, a distributed systems architecture, a robotics benchmark and controller, an automotive localization dataset, a categorical foundation-model framework, or a metaphor for traversing a complex technical landscape (Costa et al., 2024, Pu et al., 2016, Misback et al., 2023, Chatzakis et al., 2023, Kurda et al., 16 Dec 2025, Wang et al., 11 Aug 2025, Mahadevan, 25 Jun 2026). The term therefore functions as a cross-domain label whose meaning is determined by local research context: in some cases it names a concrete artifact, and in others it marks a staged or exploratory methodology (Ma, 2020, Kim et al., 2024).

1. Terminological scope and recurring uses

The term appears in at least three distinct ways. First, it can identify an institutional or experimental setting. In the speech domain, Odyssey 2024 was the Speaker and Language Recognition Workshop venue and the framework within which the Speech Emotion Recognition Challenge was organized; it was explicitly “not a standalone dataset or model” (Costa et al., 2024). Second, it can name a concrete software or data artifact, as in the CUDA-based general-relativistic radiative transfer code Odyssey (Pu et al., 2016), the floating-point rewriting workbench Odyssey (Misback et al., 2023), the distributed data-series framework Odyssey (Chatzakis et al., 2023), the resilient distributed training system Odyssey (Zhou et al., 29 Aug 2025), and the automotive LiDAR-Inertial Odometry dataset Odyssey (Kurda et al., 16 Dec 2025). Third, it can be used metaphorically, as in “Segmentation Loss Odyssey,” where the title presents the subject as a journey through the landscape of medical image segmentation losses (Ma, 2020), or in “A Data-Driven Odyssey in Solar Vehicles,” where the “Odyssey” is the long, uncertain, and strategic journey of solar-vehicle travel (Kim et al., 2024).

A separate usage belongs to planetary science, where Mars Odyssey is the mission platform rather than a methodological metaphor. In that literature, Mars Odyssey is a NASA Mars orbiter operating in polar orbit since February 2002, and the term designates the spacecraft whose Neutron Spectrometer data are used for hydrogen mapping and polar seasonal studies (Mesick et al., 2019). This suggests that “Odyssey” in technical writing frequently signals either extended traversal through a difficult state space or a platform intended to support such traversal.

2. Odyssey as a speech technology venue and challenge framework

Within speech and language processing, Odyssey most prominently denotes the Odyssey 2024 Speaker and Language Recognition Workshop and its Speech Emotion Recognition Challenge (Costa et al., 2024). The challenge had two independent tasks: a categorical emotion recognition task and a continuous emotion attribute prediction task. For the categorical task, systems classified each speech segment into one of eight emotion categories: anger, happiness, sadness, fear, surprise, contempt, disgust, and neutral. The data came from MSP-Podcast, and the official evaluation metric was Macro-F1 because the class distribution was imbalanced (Costa et al., 2024).

Two challenge systems illustrate how Odyssey functioned as a comparative evaluation framework. The “Double Multi-Head Attention Multimodal System” combined frozen self-supervised acoustic encoders, Whisper transcription, and BERT-large-uncased text representations in an early-fusion architecture with two attention stages: the first learned cross-modal contextualized representations, and the second pooled them into an utterance-level embedding (Costa et al., 2024). The authors explored wav2vec 2.0 large-lv60k, XLS-R 300M, HuBERT large, and wavLM Large as acoustic encoders, used weighted sums across all 24 Transformer layers for speech features, and reported a hard-voting ensemble that achieved 34.41% Macro-F1 on the challenge test set, ranking third among 31 participating teams in the categorical task (Costa et al., 2024).

The first-place Task 1 system addressed the same challenge through class-imbalance mitigation rather than double attention (Chen et al., 2024). It used an ensemble of 7 independently trained multimodal models, with text features derived from Whisper-large-v2 transcriptions further improved by an ASR error correction model, and Roberta-large text representations formed from the average of the last 4 layers (Chen et al., 2024). Its central optimization device was class-weighted focal loss,

$$\mathcal{L}_\text{focal}' = \frac{1}{N}\sum_{i=1}^{N} -w_j(1-p_i)^\gamma \log p_i,$$

which combined focal loss with prior-based class weights to better balance major and minor emotions (Chen et al., 2024). The final majority-voting ensemble achieved 35.69% Macro-F1 and 37.32% accuracy, ranking first among 68 submissions (Chen et al., 2024). Across these systems, Odyssey served as the experimental locus for comparing multimodal fusion, class-imbalance handling, and ensemble design under natural, spontaneous, and class-imbalanced speech conditions.

3. Odyssey as interactive and distributed systems infrastructure

Several computer-systems papers use Odyssey to denote infrastructure for diagnosis, scheduling, optimization, or recovery. In floating-point analysis, Odyssey is an interactive workbench for expert-driven floating-point expression rewriting that organizes numerical debugging around a three-stage workflow: diagnose problems, generate solutions, and tune results (Misback et al., 2023). It wraps Herbie rather than replacing it, exposes internal heuristics such as local error, supports interactive range modification, maintains a rewritings table, and enables manual and automated rewrites to be compared on the same error plots (Misback et al., 2023). In a user study with five expert numerical analysts, participants completed five out of seven challenging tasks on average in about 40 minutes after a 12-minute tutorial, and unanimously praised interactive range modification and local error visualization (Misback et al., 2023).

In large-scale data processing, Odyssey is a distributed framework for exact data-series similarity search on clusters of multi-core servers (Chatzakis et al., 2023). Its design combines intra-node parallelism, inter-node distribution, prediction-based scheduling, work stealing without moving raw data, BSF sharing, flexible partial replication, and density-aware partitioning based on Gray-code ordering of summarization buffers (Chatzakis et al., 2023). The framework targets exact 1-NN similarity search, typically under Euclidean distance,

$ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$

while also discussing extensions to $k$-NN and DTW (Chatzakis et al., 2023). Reported results include dynamic prediction-based scheduling that was up to 150% better than weaker scheduling methods in some cases, work stealing that was up to almost $2\times$ faster than the non-stealing version for large node counts under full replication, and a best algorithm up to $6.6\times$ faster than the strongest competitor (Chatzakis et al., 2023).

In federated query optimization, Odyssey is a cost-based optimizer for federated SPARQL queries that uses Characteristic Sets, Characteristic Pairs, Federated Characteristic Sets, and Federated Characteristic Pairs to improve source selection, join ordering, and query decomposition (Montoya et al., 2017). For star-shaped subqueries with predicate set $P$, the paper gives

$cardinality(P) = \sum_{C_j} count(C_j)$

for DISTINCT queries and a product-based estimate for non-DISTINCT ones (Montoya et al., 2017). On FedBench, the optimizer selected at least $1.81\times$ fewer sources than FedX and SemaGrow on average, produced at least $2.62\times$ fewer subqueries than HiBISCuS, and was on average at least $25.46\times$ faster than FedX-Warm and $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$0 faster than SPLENDID (Montoya et al., 2017).

A later distributed-training paper uses Odyssey for backup-free fault tolerance in LLM training (Zhou et al., 29 Aug 2025). There, Odyssey adaptively chooses between dynamic parallelism and data rerouting at failure time by combining a unified performance model, execution-plan search, performance estimation, and communication optimization (Zhou et al., 29 Aug 2025). The optimization objective is posed over a sequence of cluster states $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$1, with total processed samples and total time written as sums over states and state transitions (Zhou et al., 29 Aug 2025). On a 32-Ascend-910B-NPU cluster, the system maintained a post-recovery performance gap within 11.00% of fault-free training, achieved up to $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$2 and $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$3 higher average throughput than Oobleck and Recycle, respectively, and preserved convergence with no observed OOM issues (Zhou et al., 29 Aug 2025).

4. Odyssey in astrophysical computation and Mars remote sensing

In computational astrophysics, Odyssey is a public GPU-based code for general-relativistic radiative transfer in Kerr spacetime (Pu et al., 2016). It computes images, spectra, spectrograms, and light curves by coupling direct numerical integration of null geodesics to the covariant radiative transfer equation in Boyer-Lindquist coordinates (Pu et al., 2016). The code integrates six ODEs for $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$4, uses CUDA C/C++ with one CUDA thread mapped to one image pixel or ray, and solves the radiative transfer problem through decoupled ODEs for optical depth and Lorentz-invariant intensity (Pu et al., 2016). The reported single-GPU performance on an NVIDIA GeForce GTX 780 Ti exceeded 1 nanosecond per photon per Runge–Kutta integration step, and the software was released publicly together with Odyssey_Edu, a GUI-based educational tool for real-time visualization of null geodesics around a Kerr black hole (Pu et al., 2016).

Planetary-science usage refers not to a software stack but to the Mars Odyssey spacecraft and, specifically, the Mars Odyssey Neutron Spectrometer (MONS). MONS measures neutron leakage albedo produced when galactic cosmic rays strike Mars, and it collects neutron fluxes in three energy bands: thermal $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$5, epithermal $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$6, and fast $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$7 (Mesick et al., 2019). Because epithermal neutron flux is strongly suppressed by hydrogen in the near-surface layer, the instrument provides a proxy for near-subsurface water equivalent hydrogen in the top roughly meter of Martian regolith (Wilson et al., 2017).

A pixon-based reconstruction of frost-free MONS prism-1 data yielded an approximately two-fold improvement in linear spatial resolution, from a raw footprint of about $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$8 FWHM to an effective reconstructed resolution of about $ED(T,S) = \sqrt{\sum_{i=1}^{n} (t_i - s_i)^2},$9 FWHM (Wilson et al., 2017). The reconstructed map revealed $k$0 wt.% WEH on the flanks of the Tharsis Montes, $k$1 wt.% WEH at Aeolis Planum in the Medusae Fossae Formation, exceptionally dry southern Elysium Planitia with $k$2 wt.% WEH in the most neutron-bright area, and no positive correlation between recurring slope lineae and subsurface hydration; the two-sample Kolmogorov–Smirnov test for RSL and non-RSL pixels gave statistic $k$3 and $k$4 (Wilson et al., 2017). The Medusae Fossae values were interpreted as too high to be explained by ordinary hydrous silicates alone, implying bulk water ice or at least ice-rich material (Wilson et al., 2017).

A later reprocessing paper extended MONS thermal and epithermal time-series maps through December 2017, spanning 17 Earth years of operation and 8 Mars Years (Mesick et al., 2019). The new pipeline was built independently, processed more than 23 million data points, removed 14.1% of the data overall, and introduced updated gain corrections, quadratic background fitting, a different GCR normalization reference date, atmospheric corrections, and explicit ground-track registration (Mesick et al., 2019). The main science result was that Martian polar neutron signatures were highly reproducible from year to year, except for a northern winter thermal-neutron anomaly from about $k$5 in MY 28 to $k$6 in MY 29, where the count rate was about 14% lower than the average of the other years after the planet-wide dust storm near $k$7 in MY 28 (Mesick et al., 2019).

5. Odyssey in mobility, simulation, robotics, and localization datasets

Another cluster of uses concerns long-horizon mobility under environmental and sensing constraints. “A Data-Driven Odyssey in Solar Vehicles” presents a Streamlit-based simulator for long-distance solar-vehicle travel that combines Google Maps API data and Solcast weather data (Kim et al., 2024). The workflow has three stages—setting vehicle specifications and physical parameters, establishing a driving plan, and outputting and controlling driving information—and the dashboard updates the vehicle state every hour based on user-selected speed (Kim et al., 2024). The simulator models generated energy as

$k$8

and decomposes consumption into drag, rolling, gravitational, and system-loss terms (Kim et al., 2024). Validation on the World Solar Challenge route from Darwin to Adelaide, about 3,020 km over 8 days, found that the Daily Average Speed Strategy covered the longest total distance and performed best overall under the competition’s charging window of 06:30–19:00 and driving window of 08:00–17:00 (Kim et al., 2024).

In localization research, Odyssey is an automotive LiDAR-Inertial Odometry dataset designed specifically for GNSS-denied situations such as tunnels and parking garages (Kurda et al., 16 Dec 2025). It contains 12 distinct trajectories, each repeated three times, giving 36 total sequences and approximately 163 km of driving over about 4 hours (Kurda et al., 16 Dec 2025). The productive subsystem consists of an Ouster OS1 rev.7 128-layer lidar running at 10 Hz and an iNAT M300-TLE-LN1 automotive-grade MEMS-based INS providing raw IMU measurements at 300 Hz, while the reference subsystem is an iPRENA-M-II navigation-grade INS with a Ring Laser Gyroscope producing ground-truth pose and inertial estimates at 250 Hz (Kurda et al., 16 Dec 2025). The dataset is presented as the first publicly available dataset featuring a RLG-based INS, and its geodetic ground truth is provided in ETRS89/ETRF2024 (Kurda et al., 16 Dec 2025). Baseline evaluations with KISS-ICP, MAD-ICP, RKO-LIO, and OSM-ICP showed that parking garage and underground car park sequences did not significantly degrade performance, whereas tunnels—especially Tunnel1 and parts of HighwayTunnel3—were particularly challenging (Kurda et al., 16 Dec 2025).

In embodied AI and mobile manipulation, ODYSSEY is a unified framework for open-world long-horizon mobile manipulation on quadruped robots with manipulators (Wang et al., 11 Aug 2025). The pipeline combines egocentric sensing, an instance-level semantic graph, a coarse-to-fine hierarchical planner powered by GPT-4.1, fine-grained local manipulation guidance from Qwen2.5-VL-72B-Instruct, and a single unified whole-body policy that outputs 18 joint offsets for locomotion and manipulation (Wang et al., 11 Aug 2025). The system introduces the first comprehensive benchmark for long-horizon mobile manipulation, with 50 rigid objects, 15 containers, 30 articulated structures, 10 draggable items, and 10 realistic scenes including indoor homes, supermarkets, a restaurant, and outdoor courtyards with slopes and stairs (Wang et al., 11 Aug 2025). Reported long-horizon task success rates were 66.7 for IndoorCollect, 69.8 for RoomNavigation, 41.0 for CartDelivery, 44.9 for CabinetStorage, 56.7 for Restocking, 47.5 for Shopping, 63.3 for OutdoorCollect, and 46.4 for OutdoorDelivery, with 40%+ overall success on every long-horizon task (Wang et al., 11 Aug 2025). Real-world deployment on a Unitree Go2 quadruped with a 6-DoF Arx5 manipulator demonstrated sim-to-real transfer for navigation-to-pick and pick-and-place tasks, while failures remained for small objects because of end-effector tracking inaccuracies and visual perception errors (Wang et al., 11 Aug 2025).

6. Odyssey as categorical framework and survey metaphor

The most abstract use of the term is ODYSSEY, a categorical framework for constructing verifiable, local truth-preserving foundation models (Mahadevan, 25 Jun 2026). Its central principle is

$k$9

meaning that a domain is represented by a cover of overlapping local regions, each with local truth values and local logic, rather than by a single global embedding or opaque checkpoint (Mahadevan, 25 Jun 2026). A foundry in this framework is an organized sheaf of knowledge with an argumentation component; it includes a cover of local contexts, local representation families, restriction maps, gluing rules, obstruction policies, update obligations, and human-facing views (Mahadevan, 25 Jun 2026). The architecture assigns roles to Scylla, Homer, Athena, Prometheus, and Toulmin, uses a finite truth partition $2\times$0, formalizes foundry construction through Universal Foundry Learning, and employs TICKET—Topos Integration using Causal Kan Extension Transformers—for admitting external or pre-built models into durable ODYSSEY state (Mahadevan, 25 Jun 2026). The paper states that ODYSSEY is fully implemented and tested across a wide spectrum of concrete foundries, including retail/company, evaluation-harness, research-program, causal-claims, and multimodal assembly settings (Mahadevan, 25 Jun 2026).

A contrasting but related usage appears in “Segmentation Loss Odyssey,” a survey-style paper on loss functions for deep learning-based medical image segmentation (Ma, 2020). There the title is not a software name but a framing device for a systematic taxonomy of four categories: Distribution-based Loss, Region-based Loss, Boundary-based Loss, and Compound Loss (Ma, 2020). The paper presents general loss as expected risk $2\times$1, treats cross entropy and Dice as central prototypes of distribution-based and region-based objectives, connects boundary loss and Hausdorff-style losses through mismatch-region and distance-weighting interpretations, and argues that many apparently novel losses are reformulations, weightings, or combinations of a few core templates (Ma, 2020). In this usage, “Odyssey” denotes breadth and complexity rather than a deployable artifact.

Across these works, the recurrent technical pattern is staged composition. Whether the subject is a speech challenge, a numerical workbench, a distributed query engine, a Mars remote-sensing pipeline, a GNSS-denied dataset, a whole-body manipulation framework, or a sheaf-theoretic foundation-model architecture, Odyssey most often names a system in which progress depends on traversing multiple coupled modules, contexts, or decision layers rather than applying a single monolithic procedure (Costa et al., 2024, Misback et al., 2023, Chatzakis et al., 2023, Wilson et al., 2017, Kurda et al., 16 Dec 2025, Mahadevan, 25 Jun 2026).