Predictive Robot Programming: Theoretical and Experimental Analysis

Abstract

As the capabilities of manipulator robots increase, they are performing more complex tasks. The cumbersome nature of conventional programming methods limits robotic automation due to the lengthy programming time. We present a novel method for reducing the time needed to program a manipulator robot: predictive robot programming (PRP). The PRP system constructs a statistical model of the user by incorporating information from previously completed tasks. Using this model, the PRP system computes predictions about where the user will move the robot. The user can reduce programming time by allowing the PRP system to complete the task automatically. In this paper, we derive a learning algorithm that estimates the structure of continuous-density hidden Markov models from tasks the user has already completed. We analyze the performance of the PRP system on two sets of data. The first set is based on data from complex, realworld robotic tasks. We show that the PRP system is able to compute predictions for about 25% of the waypoints with a median prediction error less than 0.5% of the distance traveled during prediction. We also present laboratory experiments showing that the PRP system results in a significant reduction in programming time, with users completing simple robot-programming tasks over 30% faster when using the PRP system to compute predictions of future positions.

Scaffolding on-line segmentation of full body human motion patterns

“…To avoid this problem, temporary states were merged with the permanent state if the distance between them was sufficiently small. This approach is similar to the on-line HMM model learning of Dolan et al [16].…”

Incremental kinesthetic teaching of motion primitives using the motion refinement tube

“…The number of states is chosen from experimental experiences : HMMs with 10-20 states work well with whole body motion patterns (19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30). For discussion of the model selection problem for HMMs, refer to Billard et al (2006), Dixon et al (2004), Kulić et al (2007a), or Lee et al (2008). The generalized motion primitive (red solid line) is generated from the HMM parameters by the algorithm in Sect.…”

Online Segmentation and Clustering From Continuous Observation of Whole Body Motions

“…To avoid this problem, temporary states were merged with the permanent state if the distance between them was sufficiently small. This approach is similar to the online HMM model learning of Dixon et al [30].…”

“…The first fish model, developed by Kanso et al [8], relies on a 2-degree-of-freedom (DOF) nonreciprocal movement, namely, out-of-phase movement at the two joints, which propels the fish through a perfect fluid without the use of a Kutta condition to shed vortices. The second fish model, which has been developed by Xiong [30] (and studied by Mason [14]), has only 1 DOF in its movement and relies on vortex shedding to move through a perfect fluid (by the scallop theorem [24], a 1-DOF vehicle cannot locomote in a perfect irrotational fluid). In both models, the fish can propel and turn itself by periodically changing its shape.…”