Visual Predictive Control of Spiral Motion

Decision strategies for automated visual collision avoidance

“…Important to note that many reference colatitude angles could be chosen. A colatitude reference is selected by considering the nonlinearity of the reference position and the expectations on stability and feasibility [34]. In this way, there is better assurance the reference feature will be tracked from a control perspective.…”

“…A Sequential Quadratic Programming (SQP) approach employing an Active-Set strategy is used to solve the optimal control problem (26). ACADO also provides a MATLAB interface and optimized code for the visual predictive controller to be implemented on real platforms [33], [34]. An experimentally derived nonlinear point mass model was used to represent the vehicle dynamics.…”

“…To account for problem constraints, nonlinear model predictive control theory [30] has recently been used to derive visual predictive control [31] approaches [32]. Visibility, state and control constraints can be explicitly considered in the control design whilst guiding a safe spiral-like avoidance path [33], [34]. In most cases, static objects are considered while dynamic objects are managed under simplifying assumptions or for a subset of collision geometries.…”

Mobile Robot Navigation in a Cluttered Environment via Visual Predictive Control

“…In Flécher et al (2019), the VPC scheme is used to simultaneously control two robotic arms sharing a same workspace. In addition to robotic arms, other systems were considered such as a flying camera in Heshmati-alamdari et al (2014) and Mcfadyen et al (2014), a mobile robot in Ke et al (2017), and a fixed-wing aerial vehicle in Lee et al (2011). These works focus on the design of the VPC scheme subject to control inputs and field of view constraints, usually for small displacements of the camera between the initial and desired poses, which limits the impact of the prediction errors and the need for a large prediction horizon.…”

Reciprocal collision avoidance for quadrotors using on-board visual detection

“…The experiments have been done with only a single quadrotor avoiding static obstacles monitored by a Motion Capture System (MCS) and the visual processing carried out offboard. In a successive contribution, the authors improved their methods by optimizing the trajectories using a model predictive control method [5]. Another solution with a vision system has been proposed in [6], where the on-board camera of an AR-Drone was used to estimate the other quadrotor's position using a specific marker.…”