Vision‐based isolated translations and rotations estimation for inverted robot approaching a target

Vision‐based inverted robot control exhibits a complex and a multi‐parameter estimation task. Compromises over speed and accuracy must be made to reduce the cost of the system, if high profile techniques are not utilized. The purpose of this paper is to present such a technique where many parameter estimation problems admit partitioning process. The process eliminates the complexity by dividing such a high dimension task into several reduced dimension problems. The partition procedure is defined by human understanding of the task. Here a mechanical setup is designed that handles the pose parameters estimations as a set of sub‐problems. The estimated pose parameters are applied to an inverted robot approaching a target on the floor.

A novel technique that helps the inverted robot to approach its target is detailed here. The new methodology is based upon the modified versions of existing and proven techniques of pure translations and rotations estimations. Providing valid conditions, the parameters of the two vectors of the camera's pose are isolated. In the first pass, the roll angle is adjusted. A subsequent pass uses modified 2‐point algorithm to estimate partial translational vector. Introducing 4‐point algorithm; an extension of pure rotational vector estimation technique, to estimate partial rotational vector. Lastly, visual depth is estimated to complete the task. For simplicity, the robot dynamics are not detailed here. It is assumed that the robot can possibly achieve any position if the desired pose parameters are known.

It is found that the isolated vectors estimation process reduces the complexity of the system and so reduces the computational cost and processing time. The proposed technique is applied to a prototype inverted Cartesian robot having 3D rotary wrist. Through analysis, it is observed that the estimated parameters are very close to the actual pose parameters.

The proposed technique can aid CNC vertical milling machines to countercheck the exact position and orientation of the tool w.r.t. the job. Similar systems in practice are DECKEL MAHO, DMU 60/80/100, 5‐axis package and DMC 35 V series.

The dynamic nature of the method proposed in the paper makes it more efficacious for mechanical/robotic systems with vertically downward tool under gravity effect.