Synthesis of Natural Squat-To-Stand Motion Using Single-Term Cost Functions

By utilizing predictive simulations, it becomes possible to generate natural human movements without relying on the experimental data. Nevertheless, accurately predicting motion still presents a considerable obstacle due to the complex nature of comprehending and replicating human motor control. In this research, a muscle-driven musculoskeletal model was employed, along with a direct collocation optimization method, to synthesize the squat-to-stand motion. We employed seven different single-term cost functions during the process and compared the resulting joint angle and velocities and the muscle activation patterns, with those healthy subjects. In general, the cost function minimizing the joint forces showed the best overall match with the experimental data, resulting in root mean square errors of 2.5 degrees for the ankle, 6.1 degrees for the knee, and 21.3 degrees for the hip joints. It was concluded that with a single-term cost function, some essential aspects of the motor control strategies are missed and for more accurate predictions, employment of multi-term cost functions is inevitable.