DVS Gazebo Plugin

This package provides a DVS simulation implemented as Gazebo plugin.

Install

First, make sure the DVS datatypes are available in your installation. For this, clone the RPG DVS ROS package into your catkin workspace.

Then, clone this package into your workspace and rebuild.

Usage

This plugin can be used as a drop-in replacement for normal Gazebo camera plugins. Both, the DVS plugin and the CameraPlugin use the Gazebo CameraSensor internally.

The following SDF snippet shows an example usage:

<sensor name='camera' type='camera'>
    <camera name='__default__'>
        <horizontal_fov>1.8</horizontal_fov>
        <image>
            <width>128</width>
            <height>128</height>
        </image>
        <clip>
            <near>0.1</near>
            <far>100</far>
        </clip>
    </camera>
    <always_on>1</always_on>
    <update_rate>60</update_rate>
    <visualize>0</visualize>
    <plugin name='camera_controller' filename='libgazebo_dvs_plugin.so'>
        <cameraName>camera_front</cameraName>
        <robotNamespace>AADC_AudiTT</robotNamespace>
        <eventThreshold>10</eventThreshold>
        <cameraInfoTopicName>camera_info</cameraInfoTopicName>
        <!-- <eventsTopicName>events</eventsTopicName> -->
    </plugin>
</sensor>

The parameters robotNamespace, cameraName and eventsTopicName (default: "events") result in "$robotNamespace/$cameraName/$eventsTopicName" as the identifier of the provided events topic. In this case, events will be accessible from "/AADC_AudiTT/camera_front/events".

The parameter eventThreshold specifies the pixel-wise threshold which has to be exceeded for a event to be emitted for this pixel.

The sensor parameter update_rate has only limited effect in Gazebo. The real rate is determined by the rendering pipeline and can be way lower than the specified rate. Still, this implementation yields a higher event frequency than similar Python-based implementations as a standalone node.

Acknowledgement

If you used this code for your research, please consider citing the paper Towards a framework for end-to-end control of a simulated vehicle with spiking neural networks.

@INPROCEEDINGS{7862386,
author={J. Kaiser and J. C. V. Tieck and C. Hubschneider and P. Wolf and M. Weber and M. Hoff and A. Friedrich and K. Wojtasik and A. Roennau and R. Kohlhaas and R. Dillmann and J. M. Zöllner},
booktitle={2016 IEEE International Conference on Simulation, Modeling, and Programming for Autonomous Robots (SIMPAR)},
title={Towards a framework for end-to-end control of a simulated vehicle with spiking neural networks},
year={2016},
pages={127-134},
keywords={automobiles;cameras;complex networks;feedforward neural nets;learning (artificial intelligence);mobile robots;DVS;camera images;complex networks;deep learning architectures;end-to-end simulated vehicle control;hand-crafted feature detectors;neural self-driving vehicle applications;neurorobotics applications;rate-based neural networks;silicon retina;spiking neural networks;steering wheel decoder;vehicle end-to-end for lane following behavior;Biological neural networks;Brain modeling;Cameras;Robot sensing systems;Voltage control},
doi={10.1109/SIMPAR.2016.7862386},
month={Dec},}