Fuzzy multisensor fusion for autonomous proactive robot perception

Abstract

Abstract-Robot perception still lacks reliability in complex natural environments. A commonly used method to improve perception is to incorporate more sensors with different modalities. This leads to increased computational requirements due to the parallel processing of huge amounts of sensor data. Appropriate sensor fusion methods are needed if contradictory information is provided by different sensors. We propose a feature-based technique to fuse multimodal sensor data using fuzzy rules. Probabilistic methods are avoided by applying fuzzyfication at the feature level. We propose a higher information gain of the available sensors by utilizing robot actions to focus sensors on objects of interest. Therefore sensor readings, algorithms and robot actions are combined into feature detectors. A goaldirected activation of these feature detectors renders parallel processing of all sensor data unnecessary.Index Terms-Fuzzy behavior selection, autonomous robot, multimodal perception, active perception I. INTRODUCTION Similar to other wearable devices e.g. cell phones, PDAs, etc. robots will be ubiquitous in all kinds of environments in the near future. They will operate in natural environments that are meant to be comfortable for people. It is desirable for the robots to adapt to the environment, not vice versa. In contrast to other devices they will carry out physical tasks autonomously. Therefore perception of the environment is essential.Nowadays robots are still under development, and their perception is not very robust. Thus robots are still not able to react properly to changes in the environment. Furthermore, the level of reliability required of robots is particularly high. A success rate of 99.9% is in general not enough. Since robots will be deployed in recurring tasks, even small probabilities of failure have to be eliminated in order to provide reliable operation over long periods of time.Our goal is to make robot perception more robust and timeefficient. Their robustness will be improved by combining several sensor modalities and adopting actions to enhance perception. We propose an approach of fuzzy integration of data from different sensor modalities. This leads to appropriate processing of conflicting information and to computational efficiency.

Adaptive sensor-fusion of depth and color information for cognitive robotics

“…In our group [7] multi-sensor fusion was realized by the use of fuzzy rules. Thus, for example, tables can be found in an unknown environment.…”

Multimodal integration processes in plan-based service robot control

“…1) to reliably detect relevant parts of its natural office environment (e.g., the state of doors, detection of tables, etc.). See Weser et al [13] for more details.…”

“…In order to avoid these computational problems we propose -as published in Weser et al [13] -the practical approach to adopt a fuzzy representation of certainty about the existence and identity of objects and their defining features. An object/feature can be existent, doubtful or impossible.…”

Proactive Human Search for the Designated Person with Prior Context Information in an Undiscovered Environments

“…Recent advances of human perception have encouraged robot integration based on the multisensory data as well. Weser shows fuzzy integration of object features from multiple sensor modalities and the utilization of robot action for perception [11].…”