An algorithmic approach to some problems in terrain navigation
Artificial Intelligence - Special issue on geometric reasoning
Analog VLSI and neural systems
Analog VLSI and neural systems
Using backpropagation with temporal windows to learn the dynamics of the CMU direct-drive arm II
Advances in neural information processing systems 1
International Journal of Robotics Research
Real-time computer vision and robotics using analog VLSI circuits
Advances in neural information processing systems 2
Recent developments of electronic neural nets in North America
Journal of VLSI Signal Processing Systems - Special issue on VLSI neural networks
A multilayered neural net controller using direct learning algorithm
Computers and Electrical Engineering - Special issue on neural networks and fuzzy logic: theory and applications in robotics and manufacturing
Neural network dynamics for path planning and obstacle avoidance
Neural Networks
A discrete approach to constructive neural network learning
Neural, Parallel & Scientific Computations
Methods to speed up error back-propagation learning algorithm
ACM Computing Surveys (CSUR)
Trimming analog circuits with solid-state devices
IEEE Spectrum
Vision Chips: Implementing Vision Algorithms with Analog VLSI Circuits
Vision Chips: Implementing Vision Algorithms with Analog VLSI Circuits
Hi-index | 0.00 |
This paper presents and discusses a neuromorphic approach to designing acontrol law for locally guided robot navigation which allows forinstantaneous response to the changes in the robot environment. The controlalgorithm is defined in the discrete domain appropriate for dedicated VLSIimplementation and direct processing of discrete sensory data, such asobtained from a CCD camera. The control law is based on the principle ofvirtual force fields. The virtual forces (repulsive and circulation vectors)guiding the robot in the vicinity of obstacles are derived from the gradientfields associated with discrete representation of visual information.Discretization and computational tasks are assigned to parallel neuromorphicprocessors, which emulate the gradient operations. The particularities ofdiscrete geometrical representations of the world and the adjustable controlparameters essential for flexible and robust controller operation arediscussed in detail. Finally, a detailed scheme of a tunable navigationcontroller specifying the parameters generated internally by the network andthose that have to be provided from an external source (e.g., the operator,or a supervisory knowledge-based controller) is provided.