Q8bot: Miniature Wire-Free Quadruped
- Q8bot is a miniature, open-source quadruped that achieves dynamic locomotion through a zero-wire PCB integration and innovative design.
- It employs eight DYNAMIXEL smart servos with rigid pin-header connections to eliminate cable wear, ensuring improved robustness and reproducibility.
- Demonstrated capabilities include trotting at 5.4 body lengths per second, slope climbing, jumping, and efficient power usage for over one hour of operation.
Searching arXiv for the cited Q8bot-related papers to ground the article. Q8bot is an open-source, miniature, low-cost quadruped whose defining contribution is a zero-wire architecture in which a central printed circuit board (PCB) functions simultaneously as electrical backplane and structural spine. Reported at 220 g with an 8 cm body length, it uses eight DYNAMIXEL XL330-M077-T smart servos, supports wireless command streaming at 200 Hz, and demonstrates dynamic locomotion including walking at 5.4 body lengths per second, turning at 5 rad/s, jumping, slope climbing, and operation for over an hour on a single battery charge (Wu et al., 2 Aug 2025). In later literature, it is treated as the direct precursor of the wire-free MiNI-Q platform, which preserves the underlying assembly philosophy while addressing workspace and joint-limit constraints (Koh et al., 12 Mar 2026).
1. Research role and design objectives
Q8bot was developed for settings in which many identical legged robots must be built, deployed, and maintained efficiently, including hands-on education, outreach, swarm robotics research, and confined-space exploration (Wu et al., 2 Aug 2025). The platform emphasizes accessibility, robustness, and replicability for builders without extensive fabrication or wiring expertise. Its design target is unusual within sub-15 cm legged robotics: preserving 8-DOF agility while holding total cost to about \$300, average assembly time to under one hour, and wiring-related failure modes to a minimum.
The central design rationale is that miniature quadrupeds often trade away agility, degrees of freedom, or open-source reproducibility. Q8bot instead combines a smartphone-scale envelope with a standalone power source, dynamic locomotion, and a mechanical/electrical layout intended to reduce cable strain and connector fatigue during repetitive leg motion (Wu et al., 2 Aug 2025). This makes the platform as much a study in small-scale robotic integration as in locomotion itself.
A common misconception is that “zero wires” implies the absence of electrical interconnects. In Q8bot, the phrase refers specifically to the elimination of loose harnesses and dynamically flexing motor cables. Electrical distribution remains present, but it is consolidated into rigid PCB traces, headers, and direct actuator-to-PCB connections rather than conventional free cables (Wu et al., 2 Aug 2025).
2. Zero-wire architecture and integrated assembly
The zero-wire methodology is centered on a vertically mounted custom PCB placed along the robot’s centerline, with geometry matched to the chassis cross-section. Two identical 3D-printed main frames attach to this board via self-tapping screws, forming a rigid spine that also carries the cylindrical batteries (Wu et al., 2 Aug 2025). The PCB integrates power distribution, a boost converter, a fuel gauge, the TTL bus, and an I2C expansion port, thereby collapsing structural support and electrical routing into a single compact assembly.
Q8bot uses DYNAMIXEL XL330-M077-T smart servos, each originally designed with a three-pin TTL connector for flexible cabling. In Q8bot, all eight motor TTL cables are replaced by rigid 2.54 mm pin headers. The motors are mounted by removing their rear plates and fastening them directly to the main frames using the motors’ built-in screws; the electrical interfaces converge onto the central PCB by solder or plug-in headers (Wu et al., 2 Aug 2025). The resulting chassis is a rigid rectangular stack with no loose wires.
This architecture has several direct implications. First, the elimination of free wires removes cable snags, fatigue, and intermittent connectors under fast leg motions. Second, the common bus and pin-header scheme reduces assembly variance and wiring mistakes, improving reproducibility in classroom and laboratory settings. Third, the stack-up simplifies fabrication by relying on two identical main frames, standardized fasteners, and a consolidated PCB rather than many distinct harness and bracket subassemblies (Wu et al., 2 Aug 2025). A plausible implication is that Q8bot’s contribution lies as much in integration methodology as in its measured locomotion metrics.
3. Mechanical and electronic configuration
Q8bot’s reported hardware configuration is summarized below.
| Attribute | Reported value |
|---|---|
| Body length | 8 cm |
| Width | 7 cm |
| Standing height | 7 cm |
| Ground clearance | 5.8 cm |
| Folded volume | 12 cm × 7 cm × 5 cm |
| Mass | 220 g |
| Degrees of freedom | 8 total |
| Actuators | 8× DYNAMIXEL XL330-M077-T |
| Servo gear ratio | 77:1 |
| Rated torque | 0.23 N·m |
| Maximum speed | 48 rad/s |
| Battery capacity | 2000 mAh combined |
| Command rate | 200 Hz via ESPNow |
Each leg is a five-bar linkage with two actuators, giving four identical legs and eight total actuators (Wu et al., 2 Aug 2025). The legs use four 3D-printed segments, three small ball bearings, and three self-tapping screws per leg, with four additional screws attaching each leg to the chassis. The side-by-side paired actuator arrangement is explicitly not coaxial; the paper states that this slightly reduces theoretical reachable workspace but has minimal impact on the gait trajectories used.
Two ground-contact geometries are reported. A flat tip yields more stable sagittal motion but slower turning because the leg segments can roll in opposing directions. An offset outer segment enforces single-direction roll and noticeably improves in-place turning (Wu et al., 2 Aug 2025). Additional traction can be introduced with a rubber band or similar grippy material, especially for slope locomotion and payload transport.
The electronics are equally compact. The central PCB integrates power management, the TTL bus interface, the boost converter, the fuel gauge, and the I2C expansion port. The motors provide absolute encoders and temperature sensing, which are used in durability testing. An optional IMU can be attached through the I2C expansion port, although it was not used in the reported experiments (Wu et al., 2 Aug 2025). Wireless teleoperation and control are performed through ESPNow, with a host laptop streaming joint commands at 200 Hz.
4. Control architecture, gait synthesis, and measured locomotion
Q8bot uses a heuristic open-loop controller. Gait trajectories are precomputed on a host laptop and streamed to the robot at 200 Hz, with sinusoid amplitude, lift-to-ground ratio, and inter-leg phase offsets adjusted to produce walking, trotting, bounding, and turning (Wu et al., 2 Aug 2025). Trotting is achieved by diagonally paired legs with appropriate phase relationships. The paper gives the trajectory primitive as
No forward or inverse kinematics, Jacobians, or stability-optimal control equations are provided in the original Q8bot paper; instead, workspace analysis is illustrated graphically, and the controller remains heuristic and open-loop (Wu et al., 2 Aug 2025). This is significant because the platform’s reported performance is obtained without onboard state estimation or adaptive feedback, so the results primarily reflect mechanical design, actuator choice, and trajectory scheduling.
The reported top walking speed is 5.4 body lengths per second. With body length , this gives
consistent with the separately reported value of (Wu et al., 2 Aug 2025). Turning speed is reported as 5 rad/s. Additional demonstrated behaviors include a peak vertical jump height of 7 cm, slope climbing up to with slow trotting and low body posture, and locomotion while carrying approximately 2× body weight, specifically 440 g on a 220 g robot.
The paper also reports the Froude number,
with at maximum speed and at the endurance speed of (Wu et al., 2 Aug 2025). For endurance, the robot covered about 650 m in one hour of teleoperation at an average ground speed of while consuming 60% of the 2000 mAh battery capacity, implying well over one hour of operation on a single charge under similar usage. The minimum recorded Cost of Transport is 6.04 at 0, although the paper does not disclose the voltage and current terms needed to recompute it directly (Wu et al., 2 Aug 2025).
5. Robustness, repairability, and reproducibility
Q8bot’s robustness claims are tied closely to its assembly model. In a preliminary user study, 11 participants with varied backgrounds assembled the robot in 55.3 ± 7.3 minutes following a PDF guide; one minor soldering step for the battery clips was omitted in the timed trials and then added uniformly post hoc (Wu et al., 2 Aug 2025). The study is relevant because it operationalizes “replicability” as measured assembly effort rather than as a purely qualitative design aspiration.
Durability testing includes both cyclic actuation and impact exposure. A single-leg jumping burn-in exceeded 5000 cycles over roughly five hours, with actuator temperatures stabilizing at 37.5°C and no observed motor or gearbox damage (Wu et al., 2 Aug 2025). In desk-height drop tests onto hardwood, failures began at 75 cm with motor torque enabled for bottom or front impacts, while side, top, and tumble orientations often survived. With motor torque disabled, failures began at 100 cm for bottom or front impacts, while other orientations survived. The common failure mode was spur gear damage in the motor gearbox.
Repairability is designed into the structure. Because of the simple chassis stack-up and direct-access mounting, actuator replacement can be performed in under 2 minutes, and repairing the common failure mode requires swapping a single spur gear (Wu et al., 2 Aug 2025). The chassis itself appears more impact-resistant than the gearboxes, and torque-disable during drops reduces impact loading transmitted into the drivetrain. The robot also remained functional after more than 15 cycles of assembly and disassembly.
The open-source release includes hardware CAD, PCB design files, firmware and control code, documentation, and build instructions through a public repository, and it supports both e-commerce MJF fabrication and hobbyist FDM printing (Wu et al., 2 Aug 2025). The paper situates Q8bot comparatively as smaller than platforms such as Pupper, Petoi Bittle, and HyperDog, and much lower in cost than widely cited open-source dynamic quadrupeds such as Doggo and Mini Cheetah. It also reports a turning speed of 5 rad/s, exceeding Mini Cheetah’s reported 4.6 rad/s, while noting that its COT of 6.04 is higher than Doggo’s 3.2, which is consistent with the use of economical servos rather than torque-transparent BLDC actuation (Wu et al., 2 Aug 2025).
6. Lineage, limitations, and other uses of the name
Later work on the wire-free quadruped MiNI-Q explicitly frames Q8bot as the enabling predecessor of a newer miniature platform (Koh et al., 12 Mar 2026). In that comparison, Q8bot is described as a 220 g, wire-free, eight-actuator quadruped with parallel five-bar legs, dynamic trotting and hopping, top speed of 0.43 m/s, normalized speed of 5.38 body lengths/s, and COT 6.04. The same comparison also identifies Q8bot’s principal mechanical limitations: physical joint limits, a comparatively small reachable workspace, limited vertical travel without horizontal displacement, and exposure to hardware risk if commanded beyond hard stops. This suggests that Q8bot’s main design trade-off was efficiency and simplicity through a parallel mechanism at the cost of workspace versatility.
MiNI-Q preserves the wire-free PCB-centered assembly model but replaces Q8bot’s parallel five-bar legs with independently actuated, mechanically unbounded 2-DOF serial legs, adds a 2S battery and 6 V step-down, integrates an onboard BNO055 IMU, and redesigns the PCB for JLCPCB automated assembly (Koh et al., 12 Mar 2026). The reported consequence is a platform with slightly lower normalized top speed and somewhat higher COT than Q8bot, but with substantially expanded functional capability, including 55 mm stair climbing, 45 mm low-clearance crawling, inverted locomotion within about 0.5 s, 220 mm jumps, rotary locomotion on loose terrain, and backflips. In the Q8bot lineage, the newer platform can therefore be interpreted as a response to the specific workspace and hard-stop issues observed in the original robot.
A separate terminological issue is that “Q8bot” also appears in quantization literature as a practical label for 8-bit NLP deployment recipes rather than for the quadruped robot. In the Q8BERT context, it denotes an 8-bit BERT-based chatbot built around quantization-aware fine-tuning, with 4× model-size compression, more than 99% of BERT weights quantized in embedding and fully connected layers, and task accuracy typically within about 1% of the FP32 baseline across GLUE and 0.81% on SQuADv1.1 (Zafrir et al., 2019). In neural machine translation, the same label is used for an 8-bit inference bot that quantizes only decoding-time GEMM and GEMV operations while keeping non-linearities and outputs in FP32, yielding 4–6× decoding speed-up with BLEU and human-evaluation parity to FP32 decoding (Quinn et al., 2018). These usages are conceptually unrelated to the quadruped, except that all three invoke “Q8” to signal an 8-bit or compact design orientation.
Q8bot therefore occupies two distinct technical niches: primarily, it is a miniature wire-free quadruped for education and legged-robotics research; secondarily, it has been used as a practical shorthand for 8-bit inference deployments in NLP. In the robotics literature, however, the term most specifically denotes the 2025 open-source quadruped whose zero-wire assembly method established a reproducible template for subsequent miniature legged platforms (Wu et al., 2 Aug 2025).