技術解説 IEEE Computer

We challenged tethered, flying locusts with visual stimuli looming from the side towards one eye in a way that mimics the approach of a predatory bird. Locusts respond to the lateral approach of a looming object with steering movements and a stereotyped, rapid behaviour in which the wingbeat pattern ceases and the wings are swept into a gliding posture. This gliding behaviour may cause the locust to dive. The gliding posture is maintained for 200 ms or more after which flight is resumed with an increased wingbeat frequency or else the wings are folded. A glide begins with a strong burst of activity in the mesothoracic second tergosternal motor neuron (no. 84) on both sides of the locust. Recordings of descending contralateral movement detector (DCMD) activity in a flying locust show that it responds to small (80-mm diameter) looming stimuli during tethered flight, with a prolonged burst of spikes that tracks stimulus approach and reaches peak instantaneous frequencies as, or after, stimulus motion ceases. There is a close match between the visual stimuli that elicit a gliding behaviour and those that are effective at exciting the DCMD neuron. Wing elevation into the gliding posture occurs during a maintained burst of high frequency DCMD spikes.

Gliding behaviour elicited by lateral looming stimuli in flying locusts A Neuroethology, sensory, neural, and behavioral physiology

Hyperdimensional (HD) computing is a promising paradigm for future intelligent electronic appliances operating at low power. This paper discusses tradeoffs of selecting parameters of binary HD representations when applied to pattern recognition tasks. Particular design choices include density of representations and strategies for mapping data from the original representation. It is demonstrated that for the considered pattern recognition tasks (using synthetic and real-world data) both sparse and dense representations behave nearly identically. This paper also discusses implementation peculiarities which may favor one type of representations over the other. Finally, the capacity of representations of various densities is discussed.

Classification and Recall With Binary Hyperdimensional Computing: Tradeoffs in Choice of Density and Mapping Characteristics

In this paper, we present a new bio-inspired vision system embedded for micro-robots. The vision system takes inspiration from locusts in detecting fast approaching objects. Neurophysiological research suggested that locusts use a wide-field visual neuron called lobula giant movement detector (LGMD) to respond to imminent collisions. In this paper, we present the implementation of the selected neuron model by a low-cost ARM processor as part of a composite vision module. As the first embedded LGMD vision module fits to a micro-robot, the developed system performs all image acquisition and processing independently. The vision module is placed on top of a micro-robot to initiate obstacle avoidance behavior autonomously. Both simulation and real-world experiments were carried out to test the reliability and robustness of the vision system. The results of the experiments with different scenarios demonstrated the potential of the bio-inspired vision system as a low-cost embedded module for autonomous robots.

/pdf/bio-inspired-embedded-vision-system-for-autonomous-micro-w8axajtq9l.pdf

Bio-Inspired Embedded Vision System for Autonomous Micro-Robots: The LGMD Case

Modeling and prediction of temporal sequences is central to many signal processing and machine learning applications. Prediction based on sequence history is typically performed using parametric models, such as fixed-order Markov chains ( $n$ -grams), approximations of high-order Markov processes, such as mixed-order Markov models or mixtures of lagged bigram models, or with other machine learning techniques. This paper presents a method for sequence prediction based on sparse hyperdimensional coding of the sequence structure and describes how higher order temporal structures can be utilized in sparse coding in a balanced manner. The method is purely incremental, allowing real-time online learning and prediction with limited computational resources. Experiments with prediction of mobile phone use patterns, including the prediction of the next launched application, the next GPS location of the user, and the next artist played with the phone media player, reveal that the proposed method is able to capture the relevant variable-order structure from the sequences. In comparison with the $n$ -grams and the mixed-order Markov models, the sparse hyperdimensional predictor clearly outperforms its peers in terms of unweighted average recall and achieves an equal level of weighted average recall as the mixed-order Markov chain but without the batch training of the mixed-order model.

Sequence Prediction With Sparse Distributed Hyperdimensional Coding Applied to the Analysis of Mobile Phone Use Patterns

https://bura.brunel.ac.uk/bitstream/2438/5958/4/Fulltext.pdf

A modified model for the Lobula Giant Movement Detector and its FPGA implementation

A binary Self Organizing Map (SOM) has been designed and implemented on a Field Programmable Gate Array (FPGA) chip. A novel learning algorithm which takes binary inputs and maintains tri-state weights is presented. The binary SOM has the capability of recognizing binary input sequences after training. A novel tri-state rule is used in updating the network weights during the training phase. The rule implementation is highly suited to the FPGA architecture, and allows extremely rapid training. This architecture may be used in real-time for fast pattern clustering and classification of binary features.

/pdf/a-binary-self-organizing-map-and-its-fpga-implementation-lnusxdw28g.pdf

A binary Self-Organizing Map and its FPGA implementation

A binary Self-Organizing Map and its FPGA implementation.

The Lobula Giant Movement Detector (LGMD) is a wide-field visual neuron that is located in the Lobula layer of the Locust nervous system. The LGMD increases its firing rate in response to both the velocity of the approaching object and its proximity. It has been found that it can respond to looming stimuli very quickly and can trigger avoidance reactions whenever a rapidly approaching object is detected. It has been successfully applied in visual collision avoidance systems for vehicles and robots. This paper proposes a modified LGMD model that provides additional movement depth direction information. The proposed model retains the simplicity of the previous neural network model, adding only a few new cells. It has been tested on both simulated and recorded video data sets. The experimental results shows that the modified model can very efficiently provide stable information on the depth direction of movement.

/pdf/a-modified-neural-network-model-for-lobula-giant-movement-x56wyfmcic.pdf

A modified neural network model for Lobula Giant Movement Detector with additional depth movement feature

The Lobula Giant Movement Detector (LGMD) is a wide-field visual neuron that is located in the Lobula layer of the Locust nervous system The LGMD increases its firing rate in response to both the velocity of the approaching object and its proximity It has been found that it can respond to looming stimuli very quickly and can trigger avoidance reactions whenever a rapidly approaching object is detected It has been successfully applied in visual collision avoidance systems for vehicles and robots This paper proposes a modified LGMD model that provides additional movement depth direction information The proposed model retains the simplicity of the previous neural network model, adding only a few new cells It has been tested on both simulated and recorded video data sets The experimental results shows that the modified model can very efficiently provide stable information on the depth direction of movement

Cy Pettit

Papers

A modified model for the Lobula Giant Movement Detector and its FPGA implementation

A binary Self-Organizing Map and its FPGA implementation

A binary Self-Organizing Map and its FPGA implementation.

A modified neural network model for Lobula Giant Movement Detector with additional depth movement feature

A modified neural network model for Lobula Giant Movement Detector with additional depth movement feature.