Minho@Home

LOCOMOTION PLATFORM

The robot stands on a metal base locomotion platform and is moved by 4 omni-directional wheels perpendicularly arranged, for greater manoeuvrability and control. The robot's locomotion module includes four motors coupled with encoders, two motor controller boards (MD49), and a main robot locomotion management board.

Figure 1 – Locomotion platform

Structure of the lower limbs

The robot tries to resemble the human body, in types of segments / members as well as in terms of proportionality, being dimensioned based on a 1,60 meters tal individual. Its body is made up by segments of polypropylene type lattice structure, with cover plates of the same material for aesthetically appealing appearance.

The activation of the lower limbs is carried out by the application of tension on straps, recreating the rotational movement of the ankles, knees and basin, just like a human being. Being the shins, the thighs and the trunk, lattice blocks pivot in the connecting zones where the movement is made, a degree of freedom of rotation in each connection, simulating the movement of several leg muscles between which the twins, Femoral biceps, quadriceps and buttocks.

Figure 2 – First sketch of the robot’s structure

ARM STRUCTURE

Also, identical to a human physiognomy, the upper limbs present 2 degrees of freedom of rotation from the trunk, i.e., in the shoulder relative to the human anatomy. Recreating thus, the movement of the contraction of the deltoid muscles, pectorals, trapezius and great dorsal of the human body. The link between the forearm and the arm also presents a degree of freedom of rotation, the elbow, recreating the movement of the biceps and triceps of the human body, thus obtaining in the segments the main degrees of freedom present in the human body.

Since the arm is one of the few ways the robot uses to interact with humans and his surrounding environment, it is fundamental to bear in mind some characteristics such as:

· Versatility

· Precision

· Reliability

· Appearance

The hand consists of 5 fingers, each one containing pressure sensors at their tips to provide an additional insight when manipulating object and to allow the robot to monitor the force exercised by its fingers. To increase the grip, the tip of each finger is coated with silicone material.

The fingers are controlled through servo motors, whose function is to move nylon threads that simulate the tendons of the human arm. Each finger demands a servo motor and the rotation of the wrist requires an extra one, making a total of 6 servo motors for each hand-wrist set.

All structures are made of PLA (polylactic acid) using 3D printing technology. This method gives us the advantage of being able to produce a greater number of parts at a reduced price as well as reduce the waiting time whenever a new concept needs to be tested.

The first wrist prototype has only roll and pitch axes, but another version which is under development includes roll, pitch and yaw axes. On top of that, a new DOF is given to the thumb. These improvements allow higher levels of freedom in the movements, gives a more realistic look and significantly improves the performance of object manipulation.

Figure 3 – Forearm example

TALKING AND SPEAK RECOGNITION

It is of extreme importance in this type of robots the requirement for voice recognition, being vital in some cases to successfully perform the tasks for which it is intended. Having that in mind, the recognition and interpretation of a natural discourse makes the whole process of communication between the human and the robot easier.

Currently, CMU Sphinx tools from Carnegie Mellon University are used for speech recognition. Essential words and phrases have been added to understand different and new commands. When the recognition software finds a sequence of keywords, the robot recognizes the action given by the human and processes acts accordingly.

In order for the humans to be able to establish a dialogue with the robot, it was necessary to implement on the robot a module to allow responding through speech. For this purpose, an Emic 2 module is used, to perform text to speech conversion.

Figure 4 – Emic 2

VISION AND NECK

For robot vision a Kinect is used. With it, it is possible to obtain important data for object recognition, object distance and colour, among others.

This device is attached to the robot’s head, making it another characteristic to approach it to the human physiognomy. To attach the head to the robot body, a neck has been developed that provides three degrees of freedom, allowing a versatility of movements for image acquisition as adequate and stable as possible.

Figure 5 - Kinect

SYSTEM ARQUITECTURE

Most of the Data Processing is carried out by a set of PC, Microprocessor and Microcontrollers. The Central Computing unit chosen for this task is a MSI Cubi 2, a Mini-Pc that became very advantageous for numerous reasons namely: its small size, fast processing and integrated peripherals.

Due to the amount of different hardware involved in this type of robot, it is almost a requirement to use ROS (Robotic Operating System). ROS offers a vast set of out of the package tools.

The implementation of ROS Nodes makes the intercommunication of all the robot modules simpler and more reliable.

Figure 6 – MSI Cubi 2

LIDAR AND IMU

A LIDAR is used as a tool for localization taking into consideration the objects that surround the robot and for a better perception of those, in order to analyse and decide the best path to travel between two points.

Several inertial sensors are used in the different members of the robot, so it can have information about the position of all the robot members, such as the trunk, the arms and the legs.

Figure 7 – Lidar

Figure 8 – First modulation of the structure of the robot