Multicriteria Optimal Humanoid Robot Motion Generation

Abstract

Humanoid robots, because of their similar structure with humans, are expected to operate in hazardous and emergency environments. In order to operate in such environments, the humanoid robot must be highly autonomous, have a long operation time and take decisions based on the environment conditions. Therefore, algorithms for generating in real time the humanoid robot gait are central for development of humanoid robot. In the early works, the humanoid robot gait is generated based on the data taken from human motion (Vukobratovic et al. 1990). Most of the recent works (Roussel 1998, Silva & Machado 1998, Channon 1996) consider minimum consumed energy as a criterion for humanoid robot gait generation. Roussel (1998) considered the minimum consumed energy gait synthesis during walking. The body mass is concentrated on the hip of the biped robot. Silva & Machado (1998) considered the body link restricted to the vertical position and the body forward velocity to be constant. The consumed energy, related to the walking velocity and step length, is analyzed by Channon (1996). The distribution functions of input torque are obtained by minimizing the joint torques. In our previous works, we considered the humanoid robot gait generation during walking and going up-stairs (Capi et al. 2001) and a real time gait generation (Capi et al. 2003). In addition of minimum consumed energy (MCE) criteria, minimum torque change (MTC) (Uno et al. 1989, Nakano et al. 1999) was also considered. The results showed that MCE and MTC gaits have different advantages. Humanoid robot motion generated based on MCE criterion was very similar with that of humans. Another advantage of MCE criterion is the long operation time when the robot is actuated by a battery. On the other hand, MTC humanoid robot motion was more stable due to smooth change of torque and link accelerations. Motivated from these observations, it will be advantageous to generate the humanoid robot motion such that different criteria are satisfied. This belongs to a multiobjective optimization problem. In a multiobjective optimization problem there may not exist one solution that is the best with respect to all objectives. Usually, the aim is to determine the tradeoff surface, which is a set of nondominated solution points, known as Pareto-optimal or noninferior solutions. The multiobjective problem is almost always solved by combining the multiple objectives into one scalar objective using the weighting coefficients. Therefore, to combine different