|
|
|
Collision avoidance and a looming sensitive
neuron: Size matters but biggest is not necessarily best. Key words: locust, looming, insect, collision-avoidance, LGMD, DCMD, neuroethology.
|
|
||||||||||||||||||||||||
|
Retinally-generated saccadic suppression
of a locust looming-detector neuron: investigations using a robot locust Roger D. Santer†, Richard Stafford and
caused by eye movements from real motion occurring within the environment. During saccadic eye movements, this task is achieved by inhibitory signals of central and retinal origin that suppress the output of motion-detecting neurons. To investigate the retinally-generated component of this suppression, we used a computational model of a locust looming-detecting pathway that experiences saccadic suppression. This model received input from the camera of a mobile robot that performed simple saccade-like movements, allowing the model’s response to simplified real stimuli to be tested. Retinally-generated saccadic suppression resulted from two inhibitory mechanisms within the looming-detector’s input architecture. One mechanism fed inhibition forward through the network, inhibiting the looming-detector’s initial response to movement. The second spread inhibition laterally within the network, suppressing the looming-detector’s maintained response to movement. These mechanisms prevent a looming detector model response to whole-field visual stimuli. In the locust, this mechanism of saccadic suppression may operate in addition to centrally-generated suppression. Because lateral inhibition is a common feature of early visual processing in many organisms, we discuss whether the mechanism of retinally-generated saccadic suppression found in the locust looming-detector model may also operate in these species. Royal Society Interface article (2004): FULL TEXT
Using the locust LGMD neuron for collision avoidance in cars LOCUST Bio-inspired visual collision detection mechanism for cars: Combining insect inspired neurons to create a robust system BioSystems 87 (2007) 164–171 FULL TEXT Locusts are able to detect colliding objects, such as avian predators, using their visual lobula giant movement detector (LGMD) neuron. This neuron and the post-synaptic descending contralateral movement detector (DCMD) neuron respond to objects on a direct collision course by producing a burst of spikes which increases in frequency as the object moves closer. The mechanisms used by the neurons to produce this response are well studied and do not involve any of the complex computational tasks such as object identification, velocity estimation or path planning that traditional automotive collision detection processes use. Computer simulations of the LGMD and its input architecture have shown similar responses to visual stimuli as are shown by the LGMD of the locust, and models coupled with mobile robots have demonstrated an ability to detect and avoid collisions in a simple environment in real time. Speeds and sizes of colliding objects in automotive situations differ considerably from both natural predators of locusts and obstacles encountered by mobile robots. The use of an LGMD model in an automotive situation would require significant adaptation of the model to its intended tasks. This study uses a similar model to that of Rind and Bramwell and Blanchard et al. 1999 TEXT to assess the plausibility of using the locust LGMD as a mechanism for detecting automotive collisions. The key parameters of the model are identified and the model is adapted to the automotive environment by adjusting these parameters with genetic algorithms (GAs). A number of collision and non-collision scenes captured on video are used to evolve the model into a suitable form to detect collisions with cars. The evolved model is then tested using further video sequences to ensure that it can robustly detect collisions and not produce false alerts during non-collision sequences. In addition to looking at the ability of the LGMD model to detect car collisions the efficiency and operation of different GAs are also investigated. The lobula giant movement detector (LGMD) of locusts is a visual interneuron that responds with an increasing spike frequency to an object approaching on a direct collision course. Recent studies involving the use of LGMD models to detect car collisions showed that it could detect collisions, but the neuron produced collision alerts to non-colliding, translating, stimuli in many cases. This study presents a modified model to address these problems. It shows how the neurons pre-synaptic to the LGMD show a remarkable ability to filter images, and only colliding and translating stimuli produce excitation in the neuron. It then integrates the LGMD network with models based on the elementary movement detector (EMD) neurons from the fly visual system, which are used to analyse directional excitation patterns in the biologically filtered images. Combining the information from the LGMD neuron and four directionally sensitive neurons produces a robust collision detection system for a wide range of automotive test situations.
Recent References: Leitinger G. Pabst M. A., Rind F. C. and Simmons P.J. Differential expression of synapsin in visual neurons of the locust, Schistocerca gregaria. J. Comp. Neurol. 480 89-100 (2004).TEXT Santer, R.D., Simmons, P. J. and Rind, F.C.. Gliding behaviour elicited by lateral looming stimuli in flying locusts. J. Comp. Physiol. A 191 61-73. (2005). TEXT Santer,R.D., Yamawaki, Y., Rind F.C.and Simmons P.J.. Motor activity and trajectory control during escape jumping in the locust Locusta migratoria. J . Comp. Physiol. A 191:965-975, 2005.TEXT Santer, R.D., Simmons, P. J. and Rind, F.C.. Gliding behaviour elicited by lateral looming stimuli in flying locusts. J. Comp. Physiol. A 191 61-73. (2005).TEXT Santer R.D., Yamawaki Y., Rind F.C. and Simmons P.J.. Motor activity and trajectory control during escape jumping in the locust Locusta migratoria. J . Comp. Physiol. A 191 965-975, 2005.TEXT Cuadri J., Linan G., Stafford R., Keil M.S. and Roca E. A bioinspired collision detection algorithm for VLSI implementation. Proceedings of the SPIE. 5839 238-248. (2005) Yue S., Rind F.C., Keil M.S., Cuadri-Carvajo J. and Stafford R. A Bio-inspired collision mechanism for cars: Optimisation of a model of a locust neuron to a novel environment. J Neurocomputing. 69 1591-1598 (2006)TEXTdoi:10.1016/j.neucom.2005.06.017 Yue S and Rind F. C Collision detection in complex dynamic scenes using an LGMD-based visual neural network with feature enhancement. IEEE Transactions on Neural Networks 17 705-716 (2006).TEXT Santer R.D., Rind F. C., Stafford R. and Simmons P.J. The role of an identified looming-sensitive neuron in triggering a flying locust’s escape J Neurophysiol 95 3391-3400 (2006)TEXT. Yue S and Rind F. C Visual motion pattern extraction and fusion for collision detection in complex dynamic scenes Computer Vision and Image Understanding 104 48-60 (2006)TEXTj.biosystems.2006.09.010 Stafford R., Santer R.D.and Rind F. C. A Bio-inspired collision mechanism for cars: Combining insect inspired neurons to create a robust system. Biosystems 87 162-169 (2007).TEXT Stafford R. and Rind F.C. Data mining neural spike trains for the identification of behavioural triggers using evolutionary algorithms. Neurocomputing 70 1079-1084 (2007).TEXT Yue S. and Rind F. C. A synthetic vision system using directionally selective motion detectors to recognize collision. Artificial Life 13 93-122 (2007) 10.1162/artl.2007.13.2.93
WEBSITE LINK
Present lab members
Collaborators
|
|
