Low-power parallel algorithms for single image based obstacle avoidance in aerial robots

Abstract

Abstract-For an aerial robot, perceiving and avoiding obstacles are necessary skills to function autonomously in a cluttered unknown environment. In this work, we use a single image captured from the onboard camera as input, produce obstacle classifications, and use them to select an evasive maneuver. We present a Markov Random Field based approach that models the obstacles as a function of visual features and non-local dependencies in neighboring regions of the image. We perform efficient inference using new low-power parallel neuromorphic hardware, where belief propagation updates are done using leaky integrate and fire neurons in parallel, while consuming less than 1 W of power. In outdoor robotic experiments, our algorithm was able to consistently produce clean, accurate obstacle maps which allowed our robot to avoid a wide variety of obstacles, including trees, poles and fences.

Reactive avoidance using embedded stereo vision for MAV flight

“…Other approaches focus on doing faster avoidance by simplifying the data coming from depth sensors: such as filtering depth camera data into planes [11], or converting dense stereo into digital elevation maps [12], or even singleimage depth from training data [13]. But again, all of these approaches demonstrate a frame rate of 20-30 Hz and require substantial off-board processing, while our conversion of disparity images to U-maps requires much less computing power.…”

“…We use the version of the ellipses rotated into the current yaw angle of the robot to simplify calculations, as these are already available from Eq. (12,13).…”

Monocular Visual-Inertial SLAM-Based Collision Avoidance Strategy for Fail-Safe UAV Using Fuzzy Logic Controllers

“…A vision-based autonomous flight with a quadrotor type UAV is proposed and tested in the Google Earth virtual environment [13]. Lenz et al [21] uses a single image captured from the onboard camera as input, produce obstacle classifications, and use them to select an evasive maneuver. Mejias and Campoy [23] presents a collision avoidance approach based on omnidirectional cameras that does not require the estimation of range between two platforms to resolve a collision encounter.…”

Automatic View Finding for Drone Photography based on Image Aesthetic Evaluation

“…Vision-based navigation is widely used in the autonomous control for the robot or the automobile (Lenz et al, 2012;Bills et al, 2011). In these applications, images provide position information for the device to locate itself in the environment for path planning.…”