How do you handle non-convexity in optimization landscapes when designing control loops for legged locomotion at Foundation Robotics Labs?
