Three different neuromorphic vision sensors; a 2D smooth optical flow sensor, a 1D tracker sensor and a 1D motion sensor are modeled in a simulator. The sensors are modeled according to their spatio-temporal properties and the model is validated with experimental data obtained in previous work. The model parameters are fitted according to validation data, where a cost function is optimized with a gradient descent approach. The simulator used is the Open Dynamics Engine (ODE) and the experimental data from previous work was collected in a typical RoboCup scenario. A soccer playing robot is here used to provide reference information from a standard digital tracker system. By modeling the output from neuromorphic vision sensors, it is possible to reconstruct the experiment in the simulator. We will present simulated data from this particular experiment and compare it to the ground truth validation data. This work is important in order to reduce the time needed for real-world experiments for the set-up and optimization phase during the integration of such sensors into robotic systems. At the end we will also make an indication of the direction of future research.

The use of three types of spatio-temporal processing elements is investigated for optical sensory preprocessing in order to solve robot control problems in mobile robotics. The sensory elements are optical analog VLSI silicon retina type devices that do on-chip gradient operations and perform a current mode hysteretic winner-take-all function. Each sensor device extracts a characteristic feature from the optical input: position of highest contrast along a 1-D array, maximum speed along a 1-D array, maximum optical flow on a 2-D array. These are continuously calculated by the respective sensory devices. The sensory devices are applied in a mobile robotics application. They are used for active ball control, ball velocity prediction and active gaze-control for RoboCup Middle-Size League robots.

The Pulse Coupled neural network, PCNN, is a biologically inspired neural net and it can be used in various image analysis applications, e.g. time-critical applications in the field of image pre-processing like segmentation, filtering, etc. a VHDL implementation of the PCNN targeting FPGA was undertaken and the results presented here. The implementation contains many interesting features. By pipelining the PCNN structure a very high throughput of 55 million neuron iterations per second could be achieved. By making the coefficients re-configurable during operation, a complete recognition system could be implemented on one, or maybe two, chip(s). Reconsidering the ranges and resolutions of the constants may save a lot of hardware, since the higher resolution requires larger multipliers, adders, memories etc.

This paper addresses the issue of data analysis for real-time identification of volatile chemicals in a portable sensor unit. One targeted application is breath alcohol detection, but there are quite a few other potential applications in areas as diverse as food safety, military reconnaissance, and consumer products manufacturing. It is shown that correct odour recognition rates as high as 96% can be achieved using mixed analog and digital VLSI circuitry for on-board data preprocessing, together with pulse coupled neural network and/or principal components analysis for feature extraction. Finally, multilayer perceptron and radial basis function neural networks are used for pattern recognition.