Learning navigational maps by observing human motion patterns

Abstract

Abstract-Observing human motion patterns is informative for social robots that share the environment with people. This paper presents a methodology to allow a robot to navigate in a complex environment by observing pedestrian positional traces. A continuous probabilistic function is determined using Gaussian process learning and used to infer the direction a robot should take in different parts of the environment. The approach learns and filters noise in the data producing a smooth underlying function that yields more natural movements. Our method combines prior conventional planning strategies with most probable trajectories followed by people in a principled statistical manner, and adapts itself online as more observations become available. The use of learning methods are automatic and require minimal tuning as compared to potential fields or spline function regression. This approach is demonstrated testing in cluttered office and open forum environments using laser and vision sensing modalities. It yields paths that are similar to the expected human behaviour without any a priori knowledge of the environment or explicit programming.

Multi-modal estimation with kernel embeddings for learning motion models

“…The paper has three main contributions; 1) the application of a Gaussian mixture pre-image algorithm to obtain an estimate of the full posterior distribution from the posterior embeddings produced by the KBR algorithm; 2) A new training scheme necessary for automatic parameter estimation. 3) A novel application of multi-modal regression for the problem of estimating viable motion paths from tracks of pedestrians that illustrates the problem with unimodal posteriors [7].…”

Bootstrapping navigation and path planning using human positional traces

“…Rather than relying on human input to determine the direction of travel we advocate that equipping the robot with person detection capabilities in situ will produce indirect observations that can be exploited for effective path planning. A navigational map constructed without any prior environment information was examined in our prior work [13]. Gaussian Processes (GP) were used to learn a navigational function that describes how human motion deviates from a shortest path prior.…”

A Robot Centric Perspective on the HRI Paradigm

“…The robot stands approximately 1.5m tall and weighs approximately 70kg. Through participation in RoboCup@Home 2010 (Alempijevic et al, 2010) and 2011 and previous work (such as Caraian & Kirchner, 2010, Richards, Paul, Webb, & Kirchner, 2010and O'Callaghan, Singh, Alempijevic, & Ramos, 2011, we have developed a rich set of core capabilities: including capabilities in locomotion, navigation, path-planning, mobile manipulation, and sensing and perception. It is this platform and these capabilities that have facilitated our research probing these branches and our paradigm.…”