Real-Time Visual Tracking and Identification for a Team of Homogeneous Humanoid Robots

Abstract

The use of a team of humanoid robots to collaborate in completing a task is an increasingly important field of research. One of the challenges in achieving collaboration, is mutual identification and tracking of the robots. This work presents a real-time vision-based approach to the detection and tracking of robots of known appearance, based on the images captured by a stationary robot. A Histogram of Oriented Gradients descriptor is used to detect the robots and the robot headings are estimated by a multiclass classifier. The tracked robots report their own heading estimate from magnetometer readings. For tracking, a cost function based on position and heading is applied to each of the tracklets, and a globally optimal labeling of the detected robots is found using the Hungarian algorithm. The complete identification and tracking system was tested using two igus Humanoid Open Platform robots on a soccer field. We expect that a similar system can be used with other humanoid robots, such as Nao and DARwIn-OP.

Online visual robot tracking and identification using deep LSTM networks

“…We instead compared our approach with three commonly used baseline methods, on our own collected datasets. For each of these baselines, we first detect the robots and form tracks as described by [28]. Then we associate these tracks to the robots and perform robot identification.…”

“…Tracks are initiated and terminated as soon as the detection appears or is missed. The Kalman-HA2 method extends this with a set of heuristics to handle false positives and false negatives in an additional post-processing step, similar to [28]. The JPDA method used as the third baseline is inspired by the JPDAR method from [10], but without using the m-best approximation in order to get the full capacity of JPDA.…”

“…The authors of this work have previously used color segmentation for humanoid robot detection in the context of RoboCup soccer [27]. In another related work pertaining to humanoid robot tracking and identification, the authors exploited a Kalman filter for motion modeling, and the Hungarian method for tracking and identification [28].…”

“…First, we need to identify each detection and then we can update the tracked positions of the robots based on the identification probabilities. Note that this is an entirely different setup than what we previously proposed [28], which was tracking in the image plane, followed by identification for the existing tracks.…”

“…In the proposed architecture we seek to avoid imparting prior information into the system through the choice of a fixed motion model, like in the Kalman filter [28]. The use of a secondary network for state prediction [5] is also avoided to keep the system unified, and to only require a single loss function and training set.…”