Chapter 1 Introduction

Abstract

Mobile robots are increasingly being employed in rough, outdoor terrain for applications such as forestry, mining, search and rescue, and hazardous site inspection (Osborn 1989; Gonthier and Papadopoulos 1998; Cunningham et al. 1999; Mae et al. 2000). These applications often require robots to travel across unprepared, rugged terrain to inspect a location or transport material. In outdoor settings, a robot’s mobility is strongly influenced by the terrain geometry and physical properties (Bekker 1956; Wong 1976). For example, a robot traversing loose, sloped sand might experience substantial wheel slippage and poor mobility, whereas a robot traversing flat, firm clay might experience excellent mobility. Outdoor operation also often requires robots to rely on on-board sensors (such as rangefinders and inertial measurement units) for navigation and control. These sensors generally contain significant uncertainty and error in their measurements. Finally, outdoor applications often require robots to operate autonomously, which requires real-time decision making that is constrained by limited onboard computational resources. The combined effects of rough, varying terrain conditions, sensor measurement uncertainty and error, and limited computation make rough terrain motion planning and control challenging problems.