IEEE Robotics & Automation Magazine - December 2016 - 90

Table 1. The parameters used in the experiments.
Parameter

Value

mg res (cm # cm)

5#5

ml res (cm # cm)

3#3

mh res (cm # cm)

1#1

o radius (m)

ml dim (m # m)

3#3

mh dim (m # m)

3#3

i radius (m)

0.35

d l (m)

1.1

d w (m)

0.6

d hw (m)

0.3

d f (m)

0.6

simultaneous localization and mapping (SLAM). The latter is
performed with Hector SLAM [23], which integrates data
from a 2-D laser scanner, Hokuyo UTM-30lx, placed at the
front of RobChair. The CollabNAV system also uses the information provided by three types of global maps constructed
offline, in particular, a static metric map, a topological map,
and a semantic map. The semantic map includes the three
main types of places being navigated (corridors, rooms, and
doors) as well as a fixed area wrapping the nodes previously
defined in the topological map. The static metric map is a grid
map provided a priori to the SLAM module. A 2-D global cost
map is obtained from the a priori grid map and includes the
laser scanner information in a 3-m radius. The low- and highresolution local cost maps are local grid maps, with dimensions of 3 × 3 m, that consist of samples of the global cost map
constrained by the current robot pose. The situation awareness module is in charge of detecting all the ambiguous situations requiring user collaboration, changes in semantic (type
of place being navigated), and environmental dynamics.

Collaborative Controller
The collaborative controller consists of a decision-making
layer that depends on the user's command, the type of place
the robot is navigating, and the ambiguous situation it is
dealing with. An ambiguous situation can be defined as the
perception of a navigation event that has multiple solutions
(e.g., intersection) or none (e.g., deadlock). The collaborative
controller receives commands from two agents: the user and
the machine agent. The user sparsely issues commands with
the self-paced P300-based BCI. Two different types of commands are possible, high-level commands consisting of a set
of target goals belonging to the navigational space (e.g., WC,
EXIT, etc.) and low-level commands consisting of a set of
local steering commands, specifically FORWARD, LEFT90,
RIGHT90, and BACK. The collaborative control architecture
combines a traded control layer with a shared control layer
(see Figure 2). The former is responsible for enabling/disabling user commands, depending on situation awareness,
and the latter is a shared controller that determines the
appropriate navigation goal according to the validated user
command, situation, and place being navigated. The traded
control layer is responsible for validating the commands provided by the user, subject to a set of perceived situations and
depending on the place being navigated. User commands are
enabled by this layer according to the following set of perceived situations:
● S1: multiple goals
● S2: multiple possible directions to avoid an obstacle
● S3: deadlock
● S4: occlusion caused by environmental dynamics
● S5: providing a global goal.
The perception module includes a situation awareness
module that informs the traded controller on the occurred
event S1 to S4. If a global goal is provided by the user, RobChair stops and expects a confirmation from the user. The
traded controller module receives the user command,
expressed either as a steering command or as a high-level
global goal, the place being navigated (semantic), and the situation S1 to S4 (or none).

Lab
End

Office

2
4
C

Path Zones
1 Corridor

1
3

Start
A
2
4
B

Decision Points
2 Room

3 Obstacle 4 Door

A Forward

B Right90

C Right90

D Left90

Figure 10. The experimental results obtained for one user, according to path zone, for a navigation task in a real scenario using selfpaced BCI. Path zone 1 corresponds to corridor, path zone 2 corresponds to room, path zone 3 denotes an obstacle avoidance, and
path zone 4 denotes a door passage. The path performed by RobChair is represented in red, and laser scan data is represented in blue.

IEEE ROBOTICS & AUTOMATION MAGAZINE

December 2016

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2016