This paper tries to carefully explain and study the motion of a legged robot on varied or different terrains. It’s really important for a legged robot to carefully plan its motion which are distinctively applied to a “Six-Legged Lunar vehicle ATHLETE” and the “humanoid HRP-2” from which the paper mostly focus on, while gaited walking may suffice on easy ground, the rough and steep terrain requires a unique sequences of footsteps and postural adjustments on robot specifically adapted to the terrain’s local geometric and physical properties.
ATHLETE is large and highly mobile about 2.75m and mass of 850kg and designed with its six articulated legs, it can roll up to 10km/hr on rotating wheels over flat smooth terrain and walk carefully on fixed wheels over irregular and steep terrain. In contrast, HRP-2 or the humanoid robot is relatively light and compact, with height 1.54m and mass 58kg and is capable of moving at up to 2.5km/hr, thus, it can walk on flat surfaces, along narrow paths, and up stairs of constant height, in addition to that, it has also been demonstrated crawling through tunnels, climbing stairs while holding onto handrails and getting up while falling down. Similarly, the two are potentially capable of scrambling across rougher terrain, such as irregular terrain where a fixed gait may be insufficient. An adequate planner is needed to make this capability a reality.
Furthermore, motion planning for legged robots on rough terrain is governed largely by two interdependent constraints namely contact and equilibrium. To keep the feet, fixed wheels or other body parts (such as hands or knees) at carefully chosen set of footfalls that is called contact and to apply forces at these footfalls that exactly compensate for gravity without causing slippage is called equilibrium. The range of forces that may be applied at the footfalls without causing slippage depends on their geometry and their physical properties, both of which vary across the terrain. So every time the robot plans to step, it faces dilemma: it cannot know the constraints on its subsequent motion until it chooses a footfall, a choice that itself depends on the constraints.
To handle such predicament, the instigator makes a key design choice similar to that of introduced by Bretl and Hauser: to choose footfalls before computing motions in which first identifying a number of potentially useful footfalls across the terrain then by stance which means mapping a robot's feet to a set of footfalls. A robot can step from one stance to another if they differ by a single footfall and if they share some feasible configuration, which called a transition. The previous planners differ primarily in which part of the problem they consider first between the Motion before footfalls and Footfalls before motion. But the previous planners opt to consider the motion for quality which often generates motion that looks unnatural and inefficient, along with the reason that ATHLETE and HRP-2 have many DOF (degrees of freedom). And so the four general strategies had been used,”Record and playback”,”Warp, blend or transform”, “Model reduction” and “Bias inverse kinematic solutions.”
Moreover, the work still has many limitations particularly in considering dynamic equilibrium, closed-loop control, visual feedback and robustness to modeling errors, error recovery of deformable terrain. Despite such limitations I agree with the authors that the planning methods presented in this paper can be useful in practice.
Although a hard core topic this article was intended to as reference in the subject area of Robotics.
ReplyDelete