Climbing robots are characterized by a secure surface coupling that is designed to prevent falling. The robot coupling ability is assured by an adhesion method leading to nonlinear dynamic models with time-varying parameters that affect the robot’s mobility. Additionally, the wheel friction and the force of gravity force are also relevant issues that can compromise the climbing ability if they are not well modeled. This work presents a model-based torque controller for velocity tracking in a four-wheeled climbing robot specially designed to inspect storage tanks. The model-based controller (MPC) compensates for the effects of nonlinearities due to the forces of gravity, friction, and adhesion through the dynamic and kinematic modeling of the climbing robot. Dynamic modeling is based on the Lagrange-Euler approach, which allows a better understanding of how forces and torques affect the robot’s movement. Besides, an analysis of the interaction force between the robot and the contact surface is proposed, since this force affects the motion of the climbing robot according to spatial orientation. Finally, simulations are carried out to examine the robot’s dynamics during the climbing movement, and the MPC is validated through the redrobot simulator V-REP and practical experiments. The presented results highlight the compensation of the nonlinear effects due to the robot’s climbing motion by the proposed MPC controller.

This work proposes a novel semantic perception system based on computer vision and machine learning techniques. The main goal is to identify objects in the environment and extract their characteristics, allowing a dynamic interaction with the environment. The system is composed of a GPU processing source and a 3D vision sensor that provides RGB image and PointCloud data. The perception system is structured in three steps: Lexical Analysis, Syntax Analysis and finally an Analysis of Anticipation. The Lexical Analysis detects the actual position of the objects (or tokens) in the environment, through the combination of RGB image and PointCloud, surveying their characteristics. All information extracted from the tokens will be used to retrieve relevant features such as object velocity, acceleration and direction during the Syntax Analysis step. The anticipation step predicts future behaviors for these dynamic objects, promoting an interaction with them in terms of collisions, pull, and push actions. As a result, the proposed perception source can assign relevant information to mobile robots, not only about distances as traditional sensors, but about other environment characteristics and object behaviors. This novel perception source introduces a new class of skills to mobile robots. Experimental results obtained with a real robot are presented, showing the proposed perception source efficacy and potential.

This work presents a new concept of environment mapping aiming to identify dynamic objects in the scene, and then estimate their direction and future position. The motion analysis is performed directly in the traditional 2D costmap through three different map frames. Costmap is a traditional mapping technique that uses an occupation grid is used, where each cell has a degree of uncertainty about whether it is occupied or not. The future pose of mobile object is estimated through its motion' behavior, analyzing its previous poses. A representation map called RGB-costmap is created by using the three layers of a traditional RGB image, where each layer has a different meaning. The red layer has the estimation of future position of the object and the current position, in the green layer are added the static objects of the map and in the blue layer there is the past positions of the dynamic object.

This paper presents an omnidirectional RGB-D (RGB + Distance fusion) sensor prototype using an actuated LIDAR (Light Detection and Ranging) and an RGB camera. Besides the sensor, a novel mapping strategy is developed considering sensor scanning characteristics. The sensor can gather RGB and 3D data from any direction by toppling in 90 degrees a laser scan sensor and rotating it about its central axis. The mapping strategy is based on two environment maps, a local map for instantaneous perception, and a global map for perception memory. The 2D local map represents the surface in front of the robot and may contain RGB data, allowing environment reconstruction and human detection, similar to a sliding window that moves with a robot and stores surface data.