Kinematics-based velocity estimation of lunar rovers

Wheeled mobile robots are increasingly being utilized in unknown and dangerous situations such as lunar surface exploration. Velocity estimation is of importance for lunar rovers to autonomously navigate and successfully traverse on rough terrains. The dynamic effects occurring at the wheel- terrain interface such as wheel slips make it difficult to real-time estimate the rover´s velocity only by position-measurement sensors. Effective acceleration sensor data are hard to observe since lunar rovers usually move at a very slow speed. This paper proposes a kinematics-based method for velocity estimation of lunar rovers by using wheel encoders and gyro measurements. Treating the lunar rover with passively compliant mechanisms as a series-parallel multi-body system, we can directly build the whole kinematics model based on closed velocity chain theory. The wheel-terrain physical contact and wheel slips involving rolling slip, side slip and turning slip are also investigated. The forward velocity calculation is achieved based on solving kinematics of the rover. Low-accuracy gyro signals are modelled as a stochastic procedure and processed with the standard Kalman filter. Physical experimental results of gyro signals filtering and velocity estimation are given to show the validity and usefulness of the methods. This kinematics-based velocity estimation method can be easily performed based on on-board sensors with low system cost.