Russell Martin

All Publications

• Improving CMA-ES convergence speed, efficiency, and reliability in noisy robot optimization problems. Evolutionary computation Martin, R. M., Collins, S. H. 2026: 1-35

  Abstract

  Experimental robot optimization often requires evaluating each candidate policy for seconds to minutes. The chosen evaluation time influences optimization because of a speed-accuracy tradeoff: shorter evaluations enable faster iteration, but are also more subject to noise. Here, we introduce a supplement to the CMA-ES optimization algorithm, named Adaptive Sampling CMA-ES (AS-CMA), which assigns sampling time to candidates based on predicted sorting difficulty, aiming to achieve consistent precision. We compared AS-CMA to CMA-ES and Bayesian optimization using a range of static sampling times in four simulated cost landscapes. AS-CMA converged on 98% of all runs without adjustment to its tunable parameter, and converged 24-65% faster and with 29-76% lower total cost than each landscape's best CMA-ES static sampling time. As compared to Bayesian optimization, AS-CMA converged more efficiently and reliably in complex landscapes, while in simpler landscapes, AS-CMA was less efficient but equally reliable. We deployed AS-CMA in an exoskeleton optimization experiment and found the optimizer's behavior was consistent with expectations. These results indicate that AS-CMA can improve optimization efficiency in the presence of noise while minimally affecting optimization setup complexity and tuning requirements.

  View details for DOI 10.1162/EVCO.a.381

  View details for PubMedID 41604289

• Comparing optimized exoskeleton assistance of the hip, knee, and ankle in single and multi-joint configurations. Wearable technologies Franks, P. W., Bryan, G. M., Martin, R. M., Reyes, R., Lakmazaheri, A. C., Collins, S. H. 2021; 2: e16

  Abstract

  Exoskeletons that assist the hip, knee, and ankle joints have begun to improve human mobility, particularly by reducing the metabolic cost of walking. However, direct comparisons of optimal assistance of these joints, or their combinations, have not yet been possible. Assisting multiple joints may be more beneficial than the sum of individual effects, because muscles often span multiple joints, or less effective, because single-joint assistance can indirectly aid other joints. In this study, we used a hip-knee-ankle exoskeleton emulator paired with human-in-the-loop optimization to find single-joint, two-joint, and whole-leg assistance that maximally reduced the metabolic cost of walking. Hip-only and ankle-only assistance reduced the metabolic cost of walking by 26 and 30% relative to walking in the device unassisted, confirming that both joints are good targets for assistance (N = 3). Knee-only assistance reduced the metabolic cost of walking by 13%, demonstrating that effective knee assistance is possible (N = 3). Two-joint assistance reduced the metabolic cost of walking by between 33 and 42%, with the largest improvements coming from hip-ankle assistance (N = 3). Assisting all three joints reduced the metabolic cost of walking by 50%, showing that at least half of the metabolic energy expended during walking can be saved through exoskeleton assistance (N = 4). Changes in kinematics and muscle activity indicate that single-joint assistance indirectly assisted muscles at other joints, such that the improvement from whole-leg assistance was smaller than the sum of its single-joint parts. Exoskeletons can assist the entire limb for maximum effect, but a single well-chosen joint can be more efficient when considering additional factors such as weight and cost.

  View details for DOI 10.1017/wtc.2021.14

  View details for PubMedID 38486633

  View details for PubMedCentralID PMC10936256