Tipos de Chatarra

This section demonstrates the implemented concept of control sharing in action. Two of the experiments are referenced here in detail. The first one has been conducted with the robot InBOT and its user only, the second much more complex one involved additional persons with shopping carts. Both experiments will be accompanied by a sketch showing the individual control shares of robot and user changing over the time.

The control shares of robot and user are calculated as follows: The velocity set-point vectors of all motion behaviors are summed up in the two classes: user and robot. The task-oriented ones count for the robot (guiding) or for the user (following). The manual steering vector counts always for the user and the obstacle avoidance behaviors always count for the robot.

Experiment with InBOT and user: This paragraph presents an extract from an experiment involving

the robot InBOT and its user. The following text describes the user’s and the robot’s actions. The text is accompanied by two figures: figure 6.20 shows the map of the mock-up shop with obstacles, products, and the paths taken by robot and user. The user was tracked using the intelligent environment (see Annex C.3). Figure 6.21 shows the control shares of user and robot during the experiment for each mode of operation.

The user is asked to pick six products (indicated by the Arabic numerals) in a given order and using a given mode of operation. The length of this shopping run adds up to 60m. The mode to be used is indicated by the Roman numerals:I for guiding, II for following, and III for manual steering. Special areas of interest

– indicated by the letters – are highlighted as well. The user starts (A), ordering the robot to guide him to the products (1) and (3). On the way to (3) he recognizes the second product (2). The user diverges and goes to (2) while the (guiding) robot stops and waits at (B). The user takes the product and finally continues towards (3). When the user is again close enough to the robot, the robot proceeds guiding the user to (3). After taking product (3), the user orders the robot to follow him and heads for product (4) and (5). He crosses the hall with the robot following him. At (D) the robot cannot follow the user due to obstacles and waits for the user to return. When the user returns, they both move around a corner and a short time later around some obstacles (E). Eventually, the user stops at product (5). After taking the product, the user grabs the force

sensitive handle and manually steers the robot back to to product (6), and then through a narrow door (F)

towards the check out counter (G). The user – by order – almost hits the corner in front of the counter and then the counter itself, but the obstacle assistant clears the situation.

Looking at the control shares (Fig. 6.21) one notices that during the Guiding Mode part the robot is in charge until the user stops at (B). The control shifts to the user and the robots slows down. After the user proceeds, the control shifts back to the robot which proceeds as well. In the Following Mode part the control is with the user leading the robot like having a virtual leash. At (D) and (E) the robot is confronted with obstacles. Thus, it demands control to avoid them. In the last part (Manual Steering Mode), the control is in general with the user until he tries to pass a narrow door (F) or steers the robot into the checkout counter (G). Here again, the robot demands control to solve the situation.

Summarizing, it can be stated that the task was performed successfully and the control sharing enabled the robot to cope with the challenges of the task.

6.7. Experiments

Fig. 6.20.: Map of the mock-up supermarket section with shelves marked in light blue, obstacles in grey, and the products to be taken with Arabic numbers. The paths of the robot (red) and the user (blue) start at

(A) and end at the checkout counter (G). The modes used are indicated byI, II, III for guiding, following

and manual steering, respectively. In the Manual Steering Mode, the user is always directly behind the robot, therefore no blue line is plotted here. Additionally, special areas of interest are marked with letters. Top left a small section of the shelves can be seen.

Fig. 6.21.: Diagram showing the control shares of user (dark blue) and robot (red). The velocity of the robot (yellow) and the distance between user and robot (black). In the first part (Guiding Mode) the control is with the robot as long as the distance between robot and user is small enough. When the distance rises (the user stops at (B)) the control shifts to the user, slowing down the robot (most of the time control and velocity have the same value, hence, the yellow and red lines overlay each other). In the second part (Following Mode) the control is with the user until the robot senses a threat by obstacles ((D), (E)). Here the robot takes control to avoid the imminent collisions, resulting in some variances of the robots velocity and distance to the user. In the final part (Manual Steering Mode) it is quite similar: the user has the majority of the control until approaching and finally passing a narrow door (F) or coming too close to an obstacle (G). Here the robot again takes over a larger control share to slow down and to change the direction of the motion.

Experiment with InBOT user, and other shoppers: An extract of the second experiment referenced

here is shown in Figure 6.22 – again accompanied by a corresponding graph of the control shares in Figure 6.23.

This experiment has been much more populated, and so the figure displaying it is much more chaotic, too, showing the paths of several objects. Robot and user start at the bottom right corner. The user is instructed to get the products (1) to (5) in the following modes: 1: (G)uiding, 2: (F)ollowing, 3: (F), 4: (G), and 5: (F). During these tasks, three other – particularly reckless – shoppers (A), (B), and (C), operating ordinary shopping carts, cross the path of InBOT, forcing an reaction. The other shopping carts had actually been ETrolley and thus had an self-localization system on their own. This position information has been used during this test as the focus was on the usability and on the navigation system (especially for avoiding moving objects), not on the object tracking system. The user has been tracked using the user tracking with

onboard sensors of LAAS (see. Annex C.2).

Again, the robot succeeded in all situations, even when impaired by its user and threatened by approaching shopping trolleys and imminent collisions. The control was shifted smoothly between robot and user so that the user did not have to be bothered with saving the robot from collisions. When the situations were solved, the robot – just as intended – automatically hands back the control to the user.