---
title: Heuristic Motion Space (HMS)
url: https://www.emergentmind.com/topics/heuristic-motion-space-hms
type: topic
---

# Heuristic Motion Space (HMS)

Searching arXiv for the cited HMS-related papers to ground the article and verify metadata.
arXiv search query: "2507.14097 Heuristic Motion Space 2507.14099 2012.08892 2102.03127 2210.08640 2311.12385 2307.07857"
Heuristic Motion Space (HMS) denotes a structured motion representation in which heuristic information shapes how motion is generated, searched, sampled, or reused. In current literature, the term appears explicitly in a hierarchical motion-memory layer for underwater manipulation, and it also functions as an interpretive framework for heuristic-guided state lattices, guiding spaces, learned search heuristics, and generative human-motion representations [2507.14099, 2012.08892, 2210.08640, 2102.03127, 2507.14097]. Across these formulations, the common pattern is that an underlying motion domain—configuration states, trajectories, latent motion tokens, or sampled paths—is not treated as a neutral space: it is reduced, biased, or metrized by heuristics that privilege particular regions, directions, or motion primitives.

## 1. Conceptual scope and terminological usage

In the explicit underwater formulation, HMS is a hierarchical layer on top of a dense PRM, implemented as a repository \(\mathcal{H}\) of motion primitives \(\{M_i\}_{i=1}^n\), each associated with an uncertainty estimate \(U_i\), stored paths \(\Pi_i\), and probabilistic reuse information. The planner selects HMS “highway nodes” and cached paths to shorten subsequent searches and updates their probabilities online [2507.14099]. In the guiding-space formalism, a related abstraction is defined as a pair
\[
(f: C \longrightarrow S,\; h: S \times S \longrightarrow \Delta(C)),
\]
where \(S\) is a guiding space, \(f\) is a projection from configuration space, and \(h\) returns a biased sampling distribution over \(C\) [2210.08640]. In heuristic-guided lattice planning, the effective motion space is the subset of motion primitives that remain after heuristic pruning at each state, rather than the full lattice action set [2012.08892]. In experience-based heuristic search, the heuristic is learned over motion primitives through a Deep Q-Network and then converted into a cost-to-go estimate for Hybrid A* [2102.03127]. In generative human motion simulation, the realized motion space consists of LLM-refined task text, motion tokens in a VQ-VAE codebook, continuous joint trajectories, and evaluation metrics such as MPJPE, PA-MPJPE, and DTW [2507.14097].

Taken together, these formulations suggest that HMS is less a single standardized object than a family of constructions in which heuristics alter the operative geometry of motion. The heuristic may be semantic, geometric, probabilistic, learned, or information-theoretic; the motion space may be continuous, discrete, hybrid, or latent.

The acronym itself is overloaded in adjacent literatures. In the cited works, “HMS” also abbreviates human motion simulation [2507.14097], human motion segmentation [2107.13362, 2506.21249], and Human Mental Search [2111.10188]. In motion-planning and motion-representation discussions, however, “Heuristic Motion Space” refers specifically to heuristic structuring of an underlying motion domain rather than those separate expansions.

## 2. Mathematical structure of HMS

Different HMS realizations begin from different base spaces, but each paper makes the underlying representation explicit.

| Setting | Underlying motion space | Heuristic structure |
|---|---|---|
| Generative human motion | \(\mathbf{M} \in \mathbb{R}^{T \times 22 \times 3}\), VQ-VAE codebook \(\mathcal{E}\) | LLM prompt rewriting, normalization, MPJPE/PA-MPJPE/DTW |
| State-lattice navigation | \(s=(x,y,\theta)\), primitives \(\gamma(s,s')\) | \(h_{2D}\), direction field, primitive pruning |
| Guided sampling | \(C\) with projection to \(S\) | \((f,h)\) with \(h:S\times S\to\Delta(C)\) |
| EBHS | Motion-primitive MDP with \(Q(s,a)\) | \(h = \log_{\gamma}(Q/R_g)\,c_a\) |
| Multi-robot RRT | Joint state space up to 65D | Learned steer \(\mathcal{H}_s\) and distance \(\mathcal{H}_d\) |
| Underwater PRM-HMS | PRM graph plus repository \(\mathcal{H}\) | Highway nodes, cached paths, BN probabilities |

In generative human motion simulation, the kinematic motion representation is
\[
\mathbf{M} \in \mathbb{R}^{T \times 22 \times 3},
\]
with joint positions \(\mathbf{P}_{t,j}\in\mathbb{R}^3\). After normalization and temporal resampling, AI-enhanced, human-prompted, and ground-truth trajectories are embedded in \(\mathbb{R}^{T^\ast \times 66}\). At the latent level, MotionGPT uses a VQ-VAE tokenizer with encoder \(E\), decoder \(D\), and codebook \(\mathcal{E}=\{\mathbf{e}_1,\dots,\mathbf{e}_K\}\), with quantization
\[
\mathbf{z}_q = \arg\min_{\mathbf{e}_i \in \mathcal{E}} \|\mathbf{z}_e - \mathbf{e}_i\|_2.
\]
This yields a discrete latent motion space of motion tokens and a continuous decoded joint-space trajectory [2507.14097].

In heuristic-guided state lattices, the motion space is a graph over discretized \((x,y,\theta)\) states, with edges given by motion primitives \(\gamma(s,s')\). The effective HMS at state \(s\) is the subset of primitives whose direction aligns with a 2D heuristic direction \(\alpha(c(s))\) within threshold \(\varepsilon=\pi/4\), together with three always-retained basis primitives: “taking a step forward,” “turning in place left,” and “turning in place right” [2012.08892].

In EBHS, the relevant space is an MDP over motion primitives. With sparse reward
\[
r(s,a,s')=
\begin{cases}
R_g & \text{if } s' \in t(S_g)\\
0 & \text{otherwise},
\end{cases}
\]
the learned action value yields a heuristic
\[
h(s_{\text{Plan}})=\log_{\gamma}\!\left(\frac{Q^*(s_{\text{MDP},k-1}, a_i)}{R_g}\right)\cdot c_a.
\]
The heuristic is therefore not a geometric distance but a transformed estimate of remaining steps in motion-primitive space [2102.03127].

In joint-space multi-robot planning, the base space is the joint kinodynamic state of all robots under double integrator dynamics, reaching 65 dimensions for 16 robots. The HMS interpretation is induced by a learned distance/cost-to-go field \(\mathcal{H}_d\) and a learned steering field \(\mathcal{H}_s\), both computed from local observations and composed centrally in the joint state space [2311.12385].

The guiding-space framework abstracts all of these by separating a reduced space \(S\) from the original configuration space \(C\). What makes \(S\) an HMS-like object is not only its lower dimensionality or altered constraints, but the fact that \(h\) turns structure in \(S\) into a biased distribution over \(C\) [2210.08640].

## 3. Heuristic operators and mechanisms

The decisive feature of HMS is the mechanism by which heuristic information reshapes motion generation or search.

In generative human motion simulation, the first heuristic is semantic. GPT-4 rewrites free-form task descriptions into motion-aware prompts aligned with MotionGPT’s training vocabulary, particularly HumanML3D action phrases such as “walks at a normal pace,” “throws forward,” and “applies paint using repetitive strokes.” This semantic projection biases inference toward regions of text space from which MotionGPT has learned valid text-to-motion mappings. A second heuristic layer appears in preprocessing: root-centering, scale normalization, axis alignment, and temporal filtering constrain trajectories to a canonical motion space before comparison. A third layer is evaluative: MPJPE, PA-MPJPE, and DTW define scalar costs over trajectories and thereby induce a metric structure on motion quality [2507.14097].

In heuristic-guided lattice planning, the central mechanism is directional pruning. A 2D Dijkstra search produces a predecessor direction \(\alpha\) for each cell. During 3D expansion, every candidate primitive has a geometric direction \(\beta\). If the deviation between \((\alpha-\theta)\) and \((\beta-\theta)\) exceeds \(\varepsilon\), the primitive is pruned. The heuristic therefore acts structurally on the action set, not merely on node ordering. The result is a heuristic-restricted motion graph rather than ordinary A* on the full lattice [2012.08892].

In EBHS, the heuristic comes from experience. A DQN evaluates all outgoing actions at a planning node; those \(Q\)-values are analytically mapped into child-node heuristics through the sparse-reward relation \(Q(s,a)\approx \gamma^L R_g\). Hybrid A* still performs explicit search and collision checking, but node expansion is biased by learned value estimates over motion primitives rather than by a purely geometric heuristic [2102.03127].

In automated parking, two heuristics are run concurrently within Shared Multi-Heuristic A*: a Reeds–Shepp heuristic that models non-holonomic constraints without obstacles, and a 2D Dijkstra heuristic that models obstacles without orientation constraints. The motion space is therefore viewed through two distinct admissible relaxations, and the shared \(g\)-values allow each heuristic to propagate improvements discovered by the other [2307.07857].

In joint-space multi-robot planning, the search tree is guided by decentralized, data-driven heuristics. The distance heuristic \(\mathcal{H}_d\) re-ranks nearest-neighbor candidates, while the steer heuristic \(\mathcal{H}_s\) replaces approximate steering with learned accelerations. This induces both a heuristic metric and a heuristic flow field over the same joint motion space [2311.12385].

In underwater manipulation, the heuristic is explicitly probabilistic. For motion primitive \(M_i\), goal \(G\), and environment state \(E\),
\[
P(M_i \mid G, E)
\propto \frac{\exp(-\lambda U_i)}{1+d(M_i,G)}.
\]
Low-uncertainty, goal-proximal primitives therefore receive higher posterior weight. Because these probabilities are updated through a Bayesian Network with real-time sensor data and path outcomes, the HMS becomes context-aware rather than static [2507.14099].

## 4. Domain-specific realizations

One major HMS realization is generative and human-centered. The G-AI-HMS pipeline operates in seven stages: text prompt input, GPT-4 prompt enhancement, MotionGPT text-to-motion generation, MediaPipe pose extraction from reference video, spatial normalization and filtering, temporal alignment by resampling to \(T^\ast\), and evaluation with MPJPE, PA-MPJPE, and DTW. Its task space covers eight industrially relevant motions including walking, throwing, painting, carrying, and sitting. The resulting HMS spans task text, latent motion tokens, and continuous kinematic trajectories, with evaluation closing the loop between generated and observed motion [2507.14097].

A second realization is graph-search based. E\(^3\)MoP uses a three-layer architecture—global path planning, local path optimization, and time-optimal velocity planning—in which the global lattice planner is itself heuristic-restricted by a 2D environment-aware direction field. The global HMS defines the coarse corridor or homotopy class inside which later optimization proceeds [2012.08892]. Automated parking extends this idea through bi-directional SMHA*, hybrid continuous-discrete states, motion primitives derived from a single-track model, and adaptive motion primitive scaling near obstacles [2307.07857]. EBHS uses a related Hybrid A* substrate, but its heuristic layer is learned from driving experience rather than handcrafted from geometry [2102.03127].

A third realization is high-dimensional learned planning. In joint-space multi-robot planning, the full search occurs in a continuous-time kinodynamic state space with double integrator dynamics, pairwise collision constraints, and obstacle constraints. The learned heuristics are decentralized at the observation level but composed centrally, allowing tree growth in up to a 65-dimensional joint space [2311.12385].

A fourth realization is memory-based and explicitly named. In underwater manipulation, HMS is a repository of highway nodes and cached motion primitives layered over PRM. For each successful goal \(g_i\), the framework stores
\[
\mathrm{HMS}[g_i]=(\Pi_i,1.0),
\]
then updates path probabilities across the repository. Planning for later goals is decomposed into partial A* searches to and from an HMS approach node, with fallback to full A* when no suitable cached route exists [2507.14099].

There is also a representation-learning branch, where HMS refers not to path planning but to the construction of a latent motion space suited for segmentation. Graph-constrained data representation learning introduces an auxiliary matrix \(\tilde X\), a nonnegative dictionary \(D\), and a coding matrix \(Z\), optimizing
\[
\mathcal{L}_G (S(\tilde{X}),{G}_0) + \lambda_0\|\tilde X-DZ\|_F^2+\lambda_1\|Z\|_F^2+\lambda_2 \operatorname{tr}(Z^T L_T Z),
\]
so that the learned representation preserves local geometry and temporal smoothness while supporting clustering [2107.13362]. Temporal Rate Reduction Clustering pursues the same aim through coding-rate terms and a temporal Laplacian, learning representations that are temporally consistent and aligned with a Union-of-Subspaces structure [2506.21249]. These works suggest a broader use of HMS as a heuristic latent space for temporal motion organization rather than for explicit control.

## 5. Empirical behavior and evaluation

Representative reported outcomes show that HMS constructions are typically justified by reductions in search effort, improvements in motion fidelity, or better clustering structure [2507.14097, 2012.08892, 2102.03127, 2311.12385, 2307.07857, 2507.14099, 2107.13362, 2506.21249].

| Setting | Reported results | Interpretation |
|---|---|---|
| G-AI-HMS | AI better in 6/8 tasks by MPJPE, 7/8 by DTW; MPJPE means 0.353 vs 0.375, \(t=-4.20\), \(p<0.0001\); DTW means 61.39 vs 67.01, \(t=-5.32\), \(p<0.0001\) | LLM-enhanced prompts moved generation toward trajectories closer to human motion |
| E\(^3\)MoP | Expanded states decreased by an average of \(66.21\%\); graph size about \(33.87\%\) of baseline; more than three times faster | Heuristic pruning reduced effective lattice size |
| EBHS | Pure DQN: \(\approx 90\%\) success in UHL; EBHS: 100% success on the same failure samples; median planning time \(\approx 60\%\) lower than Hybrid A* | Learned heuristics improved speed without inheriting RL failure cases |
| Joint-space multi-robot planning | Plain RRT fails in every instance for 4 robots; learned heuristics plan up to 16 robots in 65D; BOTH improves nodes by \(67.7\%\) vs STEER and \(81.8\%\) vs DISTANCE in one regime | Heuristic fields mitigate exponential growth in joint-space search |
| Automated parking | Slot 27: expanded states 2457 vs 73; execution time 47.8 s vs 11.51 s; reverse path length 7.29 m vs 3.29 m | Multi-heuristic guidance reduces search and preserves path quality |
| Underwater PRM-HMS | At 30,000 samples: PRM + A* \(\approx 5.6457\) s, AHMP \(\approx 2.2045\) s; PRM-vs-HMS mean absolute joint errors smaller than PRM-vs-RRT across five tests | HMS reuse improves runtime while keeping PRM-like trajectories |

The representation-learning branch reports analogous gains, but in clustering rather than control. Graph-constrained representation learning improves ACC and NMI over both unsupervised and transfer-based HMS baselines on Weizmann, Keck, UT, and MAD; for example, on Weizmann it reports ACC \(0.8501\) and NMI \(0.9053\), compared with transfer-baseline ACC \(0.6391\) and NMI \(0.8599\) [2107.13362]. Temporal Rate Reduction Clustering reports state-of-the-art HMS performance on five benchmark datasets, including Weiz \(94.12/95.91\) ACC/NMI and YouTube \(97.96/98.96\), with further gains when paired with CLIP features [2506.21249]. In these cases, the “motion space” is validated by its ability to make temporal clustering more separable and temporally coherent.

## 6. Limitations, ambiguities, and future directions

A recurrent limitation is that HMS is not yet a single canonical formalism. Several papers explicitly note that they do not use the phrase “Heuristic Motion Space” and are being interpreted through that lens rather than introducing standardized HMS theory [2012.08892, 2507.14097, 2210.08640]. This suggests conceptual breadth, but it also means comparisons across papers often involve different base spaces, different heuristic roles, and different optimality criteria.

Specific limitations recur by domain. Generative human motion simulation relies on MediaPipe as ground truth proxy, defines motion kinematically rather than dynamically, and remains challenged by tool interaction and multi-phase movements [2507.14097]. Heuristic-guided primitive pruning preserves resolution completeness through invariant basis actions, but theoretical optimality is not guaranteed [2012.08892]. EBHS has no formal guarantee of heuristic admissibility or consistency under function approximation, and its experiments do not address dynamic obstacles or multi-agent traffic [2102.03127]. Guiding-space evaluation depends on the choice of target distribution \(T\) and requires KL-based density estimation that becomes difficult in high dimensions [2210.08640]. Learned decentralized heuristics for multi-robot planning depend on expert-generated data and are presently specialized to double integrator dynamics [2311.12385]. Automated parking assumes static 2D occupancy maps and low-speed kinematic modeling [2307.07857]. The underwater HMS validation is limited to precise arm movements and does not yet cover extensive object handling or whole-body manipulation [2507.14099]. Temporal Rate Reduction Clustering notes that the Union-of-Subspaces prior is less suitable for micro-scale manipulative actions in temporal action segmentation settings [2506.21249].

The future directions are correspondingly diverse. In generative human motion, proposed extensions include multimodal fine-tuning with language, motion, and biomechanical feedback, richer prompt vocabularies, hybrid prompting with expert edits, and integration with physics-based models and digital twins [2507.14097]. In lattice planning, suggested extensions include multiple heuristic paths, adaptive thresholds, and homology-aware recovery modes [2012.08892]. EBHS identifies dynamic strategic planning as a natural next step [2102.03127]. Guiding-space analysis points toward black-box learned guiding spaces, hierarchical guiding spaces, and automated design using information-theoretic evaluation [2210.08640]. Multi-robot work highlights broader dynamics classes and generalization across robot-density regimes [2311.12385]. Parking work points toward dynamic environments, partially observable maps, and additional heuristic channels [2307.07857]. Underwater manipulation explicitly targets whole-body manipulation and more complex object interaction [2507.14099].

A plausible implication is that HMS will remain a unifying concept only if future work makes the heuristic component explicit at the level of representation, search restriction, or probabilistic guidance. The current literature already shows the essential pattern: motion spaces become substantially more useful when heuristics are not external add-ons but intrinsic operators that define which motions are generated, which states are expanded, which trajectories are retained, and which latent structures are considered meaningful.

Source: https://www.emergentmind.com/topics/heuristic-motion-space-hms