In the pursuit of replicating the rapid maneuvers seen in nature, engineers have long faced the challenge of translating complex biological dynamics into robotic systems. While animals perform acceleration, braking, and turning with apparent ease, most robotic research has concentrated on steady-state locomotion, optimizing speed and energy efficiency. The development of Baleka, a bipedal robot designed specifically for rapid acceleration and gait termination, marks a significant step toward bridging this gap.
Baleka’s design was driven by trajectory optimization techniques, a method that allows engineers to determine mechanical parameters by simulating desired tasks. The robot’s performance is measured using the Vertical Agility (VA) metric, which multiplies maximum hopping height by hopping frequency. In testing, Baleka achieved a VA of 1.86 m/s in bipedal mode and 1.82 m/s in monopod mode. Remarkably, when leaping with a single leg, it reached 1.33 m/s, surpassing the human benchmark of 0.89 m/s.
Selecting the right leg morphology was a critical early step. Engineers compared series articulate linkages, common in platforms like the MIT Cheetah, with four-bar and five-bar parallel linkages used in robots such as Ghost Minitaur and ATRIAS. The five-bar linkage emerged as optimal, offering balanced load distribution, superior force amplification, and high proprioceptive sensitivity, while maintaining a compact hip design. A linkage ratio of 0.35 was chosen, based on analysis of workspace, singularity, and torque requirements.
Actuation was another key consideration. To enable virtual compliance—actively modulating elasticity—high control bandwidth was required. Quasi direct-drive actuators were selected over geared motors to achieve high mass-specific torque output, despite sacrificing some robustness and proprioceptive sensitivity. The optimization process explored combinations of leg length and gear ratio, ultimately selecting a nominal leg length of 0.5 m and a gear ratio of 5 as the fastest configuration for a 6 m sprint starting and ending at rest.
The mechanical design emphasized modularity, robustness, low friction, and minimal mass. U12 T-Motors and Matex planetary gearboxes were integrated, and stress analyses informed material choices to ensure durability under expected ground reaction forces. The completed platform weighed 14.1 kg, with leg links comprising just over 13% of body mass—comparable to the MIT Cheetah.
Baleka’s control system used Simulink Real-Time interfaced via Beckhoff Ethercat devices, running at 500 Hz. A motion capture system provided precise position and velocity data, while force sensors detected ground contact for state transitions. The robot was mounted on a mechanical boom to constrain motion in the sagittal plane during vertical agility tests.
A virtual compliance controller modeled each leg as a spring-loaded inverted pendulum (SLIP), with adjustable spring constants and damping coefficients. This controller was implemented through a Jacobian-based transformation, converting desired virtual forces into motor torques for the five-bar linkage. A sequential state machine coordinated compression, thrust, flight, and landing phases during hopping.
Experiments assessed force transparency, robustness, and agility. In force transparency tests, calculated end-effector forces matched sensor readings with an average error of 16%, attributed to joint friction and motor cogging. Jump and drop tests confirmed the platform could land safely from its maximum hopping height—0.54 m in monopod mode and 0.92 m in bipedal mode. Continuous hopping trials measured stance and aerial times, with peak ground reaction forces reaching 900 N, about 5.8 times the robot’s mass, comparable to human jumping loads.
Vertical agility tests highlighted Baleka’s competitive edge. Its bipedal VA of 1.86 m/s exceeded most existing robots, second only to the GOAT leg’s 1.88 m/s, which benefits from not supporting a boom. In single-leg mode, Baleka’s VA of 1.33 m/s outperformed humans. The platform also achieved the highest jump height among comparable robots, excluding ultra-light designs like SALTO.
Baleka demonstrates that trajectory optimization can effectively guide mechanical design for agile, transient maneuvers. Its combination of optimized linkage geometry, high-torque actuation, and precise control positions it as a leading platform for future studies in rapid acceleration and deceleration strategies.