LAURON VI: Hexapod Dynamic Locomotion

• LAURON VI is a hexapod platform known for its dynamic locomotion, blending innovative mechanical design with advanced control strategies.
• Its design incorporates series elastic actuators, topology-optimized chassis, and three control methods (kinematic, MPC, RL) for reliable performance.
• Empirical evaluations reveal robust mixed-terrain navigation, efficient payload delivery, and adaptability for exploration and disaster-response tasks.

LAURON VI is a six-legged (hexapod) robotic platform designed for dynamic walking research, addressing the demands of traversing both highly challenging and mixed-terrain environments with rapid, reliable, and energy-efficient locomotion. Its development is motivated by the inherent stability and flexibility of hexapod systems, which traditionally have lacked fast gaits on benign surfaces, thus limiting their broader practical use. LAURON VI introduces architectural innovations and advanced control strategies, resulting in improved adaptability and performance for a variety of field applications including exploration, search-and-rescue, and mixed-terrain industrial tasks.

1. Mechanical and Structural Design

LAURON VI features an insectoid kinematic leg structure, where each of the six legs implements a 45° roll angle and a –15° pitch for enhanced continuous workspace. The design fans the front and hind legs with a yaw of ±40°, reducing workspace overlap and maximizing reach. Each leg consists of three revolute joints (alpha, beta, gamma), actuated by a total of 18 series elastic joint actuators. These actuators—adapted from ANYbotics platforms—deliver high-frequency interfaces necessary for both Cartesian impedance and pure torque control, providing compliance for shock absorption and the torque capacity required for dynamic gaits.

The robot’s main body is generated via topology optimization, resulting in a truss-like configuration constructed from carbon fiber tubes linked by custom components. This topology ensures a robust and lightweight chassis, essential for maintaining impact resistance and payload capability while minimizing mass. Additional mechanical robustness is achieved through hot-swappable 22.2 V LiPo batteries, supporting operational longevity and rapid field turnaround.

2. Control Architectures

Three distinct control approaches are implemented and evaluated on LAURON VI:

Kinematic-Based Controller

• Periodic Cartesian foot trajectories are generated, following Bézier curves for the swing phase to guarantee smooth transitions and shock absorption.
• In the stance phase, feet track inverse trajectories relative to desired body motion, supplemented by gravity compensation using feed-forward forces.
• Target joint torques are derived from analytical inverse kinematics and the leg Jacobian; local foot compliance is governed by cascaded position–velocity controllers with P (spring) and D (damper) gains inherent to the ANYdrive actuators.

Model-Predictive Controller (MPC)

• The robot’s movement is optimized by computing ground reaction forces (GRFs) over a short prediction horizon (5 steps, ~0.5 s at 100 Hz), subject to actuator and friction constraints.
• Dynamics are modeled with a Single Rigid-Body Dynamics (SRBD) formulation, where the state vector comprises Euler angles (ZYX: yaw ψ, pitch θ, roll φ), base position (p), angular velocities (ω), and linear velocities (ṗ). The key system equation is:

$$\dot{x}(t) = A_c(\psi)x(t) + B_c(r_0,...,r_5,\psi)u(t)$$

with $f_{grf,j}$ denoting the GRF at leg $j$.

• The cost function incorporates state and control penalties, and the problem is solved via quadratic programming (using, e.g., OSQP).

Reinforcement-Learned Controller

• Policies are trained in simulation via the legged_gym framework, initially on flat terrain.
• Observations include base linear acceleration, gyroscopic data, an estimated gravity vector, commanded velocities, joint positions, and action histories.
• The reward structure encompasses terms for target tracking, actuator smoothness, and symmetrical tripod gait reinforcement.
• A dedicated “Actuator Network,” trained on recorded hardware data, is integrated with additional low-pass filtering to mitigate sim2real discrepancies and stabilize actuation.

3. Empirical Evaluation and Field Deployment

Extensive laboratory and field testing validate LAURON VI’s operational envelope.

• Laboratory Trials: All three controllers were systematically evaluated for forward, backward, sideways, and rotational motion. Maximum observed forward velocities were 0.43 m/s for the RL controller, 0.17 m/s for kinematic-based, with MPC in between. The robot demonstrated stable operation under augmented payloads (additional 12 kg) and while dragging heavy objects (e.g., 22 kg boxes); MPC with an added PD compensator addressed model mismatches (notably, inaccurate center-of-mass).
• Endurance: Continuous walking was sustained for 108 minutes using hot-swappable batteries until cell low-voltage cutoff, with the robot traversing repeated 5 m courses.
• Mars Analog Mission: Deployed in the Tabernas Desert (Spain), LAURON VI negotiated diverse terrain features (flat, inclines, saline crusts, obstacles) using the kinematic controller. The robot’s adaptive stepping and body-height adjustment validated its design for real-world environment tasks. Minor mechanical issues (such as carbon-fiber tube twisting) were observed but did not impede task completion.

4. Practical Applications and Implications

The integration of dynamic gait synthesis and rapid locomotion significantly extends the operational scenarios for hexapod robots, which have historically been constrained by slower speed profiles compared to quadrupeds.

• Mixed-Terrain Navigation: Seamless transitions between dynamic (RL/MPC-driven) and stable (kinematic-based) gaits enable deployment in environments with variable ground conditions—a requirement for exploration, disaster response, and industrial inspection tasks.
• Efficient Coverage of Flat Terrain: The RL policy’s top speed (0.43 m/s) increases efficiency in missions with large extents of even ground, crucial for field logistics or area monitoring.
• Payload Delivery and Sensor Operation: The optimized structure and power subsystem support substantial payloads (computational/sensing) and extended field deployment.

These attributes collectively expand the feasible domain for autonomous hexapods, directly countering prior limitations in speed and adaptability.

5. Future Research and Development

Planned advancements for LAURON VI focus on further augmenting perceptual, control, and autonomy capabilities:

• Sensor Augmentation: Integration of proprioceptive and exteroceptive modalities, notably six depth cameras, to enhance both state estimation and environmental awareness—feeding into control pipelines such as MPC for more robust rough-terrain operation.
• Advanced Gait Synthesis: Ongoing reinforcement learning research aims to develop novel gait patterns leveraging the legged system's full kinematic and dynamic richness, moving beyond conventional tripod gaits.
• Autonomous Field Operations: The platform is intended for heterogeneous team deployments, with efforts directed toward tighter coupling with perception and autonomy frameworks for more complex mission scenarios.
• Model and Controller Refinement: Improvements to SRBD modeling, particularly for adaptive compensation of center-of-mass shifts and structural deflections, are under consideration for further enhancing predictive control performance and robustness.

6. Summary and Outlook

LAURON VI exemplifies a modern approach to hexapod robot design, pairing innovative mechanical engineering—including workspace optimization and series elastic actuation—with a spectrum of advanced control architectures. Demonstrated across both laboratory and Mars analog environments, its ability to combine speed, stability, compliance, and adaptability addresses longstanding gaps in the operational spectrum of six-legged robots. The ongoing enhancements in sensing, control, and autonomy position LAURON VI as a versatile experimental platform for advancing dynamic legged locomotion research and field robotics applications.