Anya Lynn Tascillo

An SPN-Neural Planning Methodology for Coordination of two Robotic Hands with Constrained Placement

This paper presents a planning methodology based on Stochastic Petri Nets (SPNs) and Neural nets for coordination of two robotic arms working in a space with constrained placement. The SPN planning method generates a global plan based on the states of the elements of the Universe of Discourse. The plan includes all the possible conflict-free planning paths to achieve the goals under constraints, such as specific locations on which objects have to be placed, order of placement, etc. An associated neural network is used to search the vectors of markings generated by the SPN reachability graph for the appropriate selection of plans. Moreover, it preserves all the interesting features of the SPN model, such as synchronization, parallelism, concurrency and timing of events. The coordination of two robotic arms is used as an illustrative example for the proposed planning method, in a UD space where the location of objects placement are restricted.

An in-vehicle virtual driving assistant using neural networks

A methodology has been developed that aids drivers by suggesting a safer following distance, through the use of sensors, and optionally, vehicle to vehicle communication. Given the restricted case where there is no option to swerve into another lane, a Matlab Simulink model varies vehicle dynamics, driver reaction delay, following distance, and initial speeds when a lead vehicle suddenly decelerates. Based upon the likelihood of collision, neural networks suggest a best following distance, and the benefits of reducing reaction delay with adaptive agents are quantified.

Predicting a vehicle or pedestrian's next move with neural networks

For a given digitized view of a driving scenario, motion clusters are formed, indicating possible moving object threats to the driver. Recurrent neural networks minimize hopping between clusters and predict a clusters next location. Frequency analysis then categorizes as significant motion, and then as either head-on/away or transverse motion.

A hierarchical anticipatory neural controller with fuzzy spectral filter diagnostics

A full state feedback recurrent (FSFER) neural network architecture is developed as a best representation in both the time and frequency domains for engine and chassis dynamometer modelling and control. In order to reduce the lag experienced by current robotic driver controllers, a fuzzy spectral filter is combined with radial basis function neural networks to suggest a best time to apply a throttle or brake input before velocity error feedback is available.

A hybrid approach to predictive modelling and control of automobiles in a noisy and variable environment

A hybrid control scheme is proposed that combines the advantages of various artificial intelligence technologies to better model and control transient behavior of a nonlinear system, even as its parameters are modified, via the ability to extract the cause of a change in a systems outputs. Appropriate techniques are developed for all stages of the development process, in an effort to reduce the amount of recalibration necessary by using a similar approach each time.

Neural network isolation of system inputs for transient modelling and control

A neural network is used to predict the sensitivity of a complex nonlinear system (such as an automobile) to input variation, which will aid greatly in the effort to model the system and the effects of changes to its controllers. A blend of signal processing techniques is used to provide maximum resolution neural network inputs for various drivers, vehicles, engine technologies, transmissions, velocity traces, and operating temperatures. The neural net predicts what four different vehicle outputs will be, given a sample of driving inputs.