Visually stimulated motor control for a robot with a pair of LGMD visual neural networks

Yue, Shigang; Rind, F. Claire

doi:10.1504/ijamechs.2012.052219

Cited by 6 publications

(12 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More specifically, both locusts [10,11,12,13] and flies [14] possess the amazing ability to perceive impending collisions in complex and dynamic scenes. These biological visual systems provide inspiration to computational modellers to build neuromorphic collision detectors in mobile robotic applications [15,16,17,18,19,20,21], ground vehicles [22,23,24,25] and UAVs [26,7].…”

Section: Introductionmentioning

confidence: 99%

Shaping the collision selectivity in a looming sensitive neuron model with parallel ON and OFF pathways and spike frequency adaptation

Peng

et al. 2018

Neural Networks

Self Cite

View full text Add to dashboard Cite

Shaping the collision selectivity in vision-based artificial collision-detecting systems is still an open challenge. This paper presents a novel neuron model of a locust looming detector, i.e. the lobula giant movement detector (LGMD1), in order to provide effective solutions to enhance the collision selectivity of looming objects over other visual challenges. We propose an approach to model the biologically plausible mechanisms of ON and OFF pathways and a biophysical mechanism of spike frequency adaptation (SFA) in the proposed LGMD1 visual neural network. The ON and OFF pathways can separate both dark and light looming features for parallel spatiotemporal computations. This works effectively on perceiving a potential collision from dark or light objects that approach; such a bio-plausible structure can also separate LGMD1's collision selectivity to its neighbouring looming detector - the LGMD2. The SFA mechanism can enhance the LGMD1's collision selectivity to approaching objects rather than receding and translating stimuli, which is a significant improvement compared with similar LGMD1 neuron models. The proposed framework has been tested using off-line tests of synthetic and real-world stimuli, as well as on-line bio-robotic tests. The enhanced collision selectivity of the proposed model has been validated in systematic experiments. The computational simplicity and robustness of this work have also been verified by the bio-robotic tests, which demonstrates potential in building neuromorphic sensors for collision detection in both a fast and reliable manner.

show abstract

Section: Introductionmentioning

confidence: 99%

Shaping the collision selectivity in a looming sensitive neuron model with parallel ON and OFF pathways and spike frequency adaptation

Peng

et al. 2018

Neural Networks

Self Cite

View full text Add to dashboard Cite

show abstract

“…The potential is then exponentially mapped to the firing rate (FR) at the firing cell (Equation 10), rather than a sigmoid function [14][15][16].…”

Section: Membrane Potential To Firing Ratementioning

confidence: 99%

“…In locusts, it is true that both of them cooperate well for the same goal of collision detection. In fact, some co mputational LGMD1 models have been well used in vehicles and robots for navigation and path exp loring [14][15][16]. However, some defects still remain unsolved until now.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Modelling LGMD2 visual neuron system

Yue

2015

2015 IEEE 25th International Workshop on Machine Learning for Signal Processing (MLSP)

Self Cite

View full text Add to dashboard Cite

Two Lobula Giant Movement Detectors (LGMDs) have been identified in the lobula region of the locust visual system: LGMD1 and LGM D2.LGMD1 had been successfully used in robot navigation to avoid impending collision.LGM D2 also responds to looming stimu li in depth, and shares most the same properties with LGM D1; however, LGMD2 has its specific co llision selective responds when dealing with different visual stimu lus. Therefore, in this paper, we p ropose a novel way to model LGMD2, in o rder to emulate its predicted bio-functions, moreover, to solve some defects of previous LGMD1 co mputational models. The mechanism o f ON and OFF cells, as well as bioinspired nonlinear functions, are introduced in our model, to achieve LGM D2's collision selectivity. Our model has been tested by a miniature mobile robot in real time. The results suggested this model has an ideal performance in both software and hardware for collision recognition.

show abstract

“…As human perception are environment-centralized [18] and evolved to cope with dynamic challenges. In biological visual systems, visual pathways are specialized in dealing with the motion cues and optical 'changes' [19] [20][21] [22]. Hereby we consider fusion motion features onto the multi-channels saliency map to facilitate fast response of the attention model to motion cues in an interactive environment.…”

Section: Introductionmentioning

confidence: 99%

Fusion of Saliency Maps for Visual Attention Selection in Dynamic Scenes

Xu¹,

Yue²

2013

IJARAI

Self Cite

View full text Add to dashboard Cite

Abstract-Human vision system can optionally process the visual information and adjust the contradiction between the limited resources and the huge visual information. Building attention models similar to human visual attention system should be very beneficial to computer vision and machine intelligence; meanwhile, it has been a challenging task due to the complexity of human brain and limited understanding of the mechanisms underlying the human attention system. Previous studies emphasized on static attention, however the motion features, which are playing key roles in human attention system intuitively, have not been well integrated into the previous models. Motion features such as motion direction are assumed to be processed within the dorsal visual and the dorsal auditory pathways and there is no systematic approach to extract the motion cues well so far. In this paper, we proposed a generic Global Attention Model (GAM) system based on visual attention analysis. The computational saliency map is superimposed by a set of saliency maps via different predefined approaches. We added three saliencies maps up together to reflect dominant motion features into the attention model, i.e., the fused saliency map at each frame is adjusted by the top-down, static and motion saliency maps. By doing this, the proposed attention model accommodating motion feature into the system so that it can responds to real visual events in a manner similar to the human visual attention system in a realistic circumstance. The visual challenges used in our experiments are selected from the benchmark video sequences. We tested the GAM on several dynamic scenes, such as traffic artery, parachuter landing and surfing, with high speed and cluttered background. The experiment results showed the GAM system demonstrated high robustness and real-time ability under complex dynamic scenes. Extensive evaluations based on comparisons with other approaches of the attention model results have verified the effectiveness of the proposed system.

show abstract

Visually stimulated motor control for a robot with a pair of LGMD visual neural networks

Cited by 6 publications

References 39 publications

Shaping the collision selectivity in a looming sensitive neuron model with parallel ON and OFF pathways and spike frequency adaptation

Shaping the collision selectivity in a looming sensitive neuron model with parallel ON and OFF pathways and spike frequency adaptation

Modelling LGMD2 visual neuron system

Fusion of Saliency Maps for Visual Attention Selection in Dynamic Scenes

Contact Info

Product

Resources

About