Fast and Adaptive Hopping Height Control of Single-legged Robot

Research on running robots has generally focussed on the steady-state. When the ground has limited foot placement surfaces or there are sudden changes in height then steady-state running is not possible. It becomes necessary to make step-bystep adjustments to place the foot. In this paper a mass-spring-damper model of a robot’s leg is used to develop a hopping controller capable of meeting rapid changes in demand height or flight time. Analysis of the model provides a simple method to select control parameters for effective height control without tuning or iteration. Additionally, a simple adaptive algorithm is introduced and demonstrated in simulation. The adaptive control algorithm allows rapid changes of height even when ground characteristics change. Experimental validation is ongoing and some preliminary results are provided. INTRODUCTION Mobility in robots is mostly achieved using wheels or tracks. However if a mobile robot has to be designed for locomotion over rough and unstructured terrains these technologies suffer from inherent limitations. This has motivated research into bio-inspired legged locomotion [1], [2]. Robots such as Boston Dynamics’ BigDog are now famous for their somewhat unsettling ability to traverse rough terrain and maintain balance [3]. These successful robots are the direct result of an approach to the problem of legged robot locomotion pioneered by Marc Raibert at MIT in the 1980s [4]. Raibert began by tackling the problem of a single-legged hopping robot [5]. His solution was to decouple the problem into a three part controller: height, speed and body angle. The single-legged controller was then successfully extended for bipeds and quadrupeds. The single-legged case is an important one as it provides a relatively simple platform to develop control techniques for application on multi-legged robots. Most research published on single-legged robot control has however focussed on the steady-state running case with variations in leg design and actuation methods [6]. Steady-state running is not possible when the ground has limited foot placement surfaces or sudden changes in height. In that case it becomes necessary for a single-legged hopping robot to land its foot from one spot to the next. The focus of this paper is controlling the hopping height, one of the three parts of a single-legged robot controller, to meet sudden changes in demand. The problem of foot placement has previously been addressed in [7]. Two methods for varying the ‘step length’, the length from one foot placement spot to the next, were investigated using a 2D planar hopping model: • Maintaining a constant forward speed and varying the hopping height from one step to the next. • Maintaining a constant hopping height and varying the forward speed. The height variation method used in [7] was found to produce less accurate foot placement than their alternative forward speed variation method. In this paper we contribute a height control algorithm based on an analysis of a mass-springdamper model of the hopping leg. This provides an insight into Proceedings of the ASME 2012 11th Biennial Conference on Engineering Systems Design and Analysis ESDA2012 July 2-4, 2012, Nantes, France