Comparison between gaze based and gaze-brain based navigations 

Get Complete Project Material File(s) Now! »

Haptic Obstacle Avoidance for Intuitive Powered Wheelchair Navigation

Vander poorten et al. [6] describes how through haptic feedback a fast ’bilateral’ com-munication channel is established between wheelchair controller and user and how this can lead to more intuitive and safer wheelchair navigation; By setting up a fast, bilateral communication channel between the user and the wheelchair controller, control can be shared deeper and problems such as mode confusion can be avoided [10]. The user can directly negotiate with the wheelchair controller over the haptic channel and is given the final word, as he/she can overrule unwanted wheelchair actions. A novel haptic collision avoidance scheme is proposed to demonstrate the potential of this new technology which is based on the rendering of a local map from onboard sensors. Within this environment, a set of circular collision-free paths is calculated with circular paths corresponding to a certain combination of wheelchair linear and rotational velocities (v, w). A path is then drawn starting from the wheelchair’s local coordinate system and runs until it reaches obstacles (see Figure 1.1).
Based on the previous operation, a haptic collision and obstacle avoidance algorithm was set up. for the former, When approaching an obstacle along a certain circular path, the repulsive force will push the joystick and tries to move it towards the origin along a straight line with constant angle. As a result the wheelchair will be slowed down along the same circular path. If tuned well the wheelchair should come to a rest in front of the obstacle. Haptic obstacle avoidance actively redirects the wheelchair towards circular paths with longer collision-free lengths. In addition to a radial force component that slows the wheelchair down nearing objects, the scheme foresees a tangential force com-ponent that bends wheelchair motion towards longer collision-free paths. To validate their approach, some first experiments were conducted with the haptic obstacle avoid-ance scheme. With cardboard boxes an artificial environment was built that represented an elevator. The user was asked to manoeuvre the wheelchair backwards inside this ele-vator beginning from a fixed start position solely using the haptic guidance and without looking backwards. While performing such a manoeuver without looking backwards was close to impossible if no haptic guidance was present, thanks to the haptic guidance the user managed to successfully drive into the elevator in 6 of the 10 times. Although some more experiments need to be done, those results are quite encouraging.

GPS and fuzzy logic map matching for wheelchair navigation

Ming Ren et al. [8], suggest that an essential process in wheelchair navigation is matching the position obtained from GPS or other sensors on a sidewalk network. This process of map matching in turn assists in making decisions under uncertainty. However, GPS-based wheelchair navigation systems have difficulties in tracking wheelchairs in urban areas due to poor satellite availability. To overcome this, a fuzzy logic-based algorithm is applied to effectively perform matching wheelchair movements on sidewalks. Fuzzy logic, based on fuzzy reasoning concepts, in many circumstances can take noisy and imprecise input to yield numerically accurate output. To validate its reliability, experiments were carried on. Map-matching GPS points on a map is twofold: (1) identification of the correct segment and (2) determination of the vehicle location on the selected segment. In this algorithm, a three-stage process was used for finding the correct segment. The process has the following steps: (1) the initial map-matching process, (2) the subsequent map-matching process along a segment, and (3) the renewed map matching across an intersection. The flowchart of the fuzzy logic maps-matching process can be found in the figure (1.2) In order to validate the fuzzy logic map-matching algorithm, the process of fuzzy logic map matching is implemented and tested on a sidewalk network. To test performances of the map matching in terms of accuracy and consuming time, further analysis is done by matching GPS points on three routes. The map-matching performances on three chosen routes are presented in Table 1.1. Testing results show that both positioning systems and the density of the sidewalk network affect the accuracy of this map-matching algorithm.The accuracy in the experiments is mainly influenced by failure in differentiating two sides of a road by only using stand-alone GPS data, so the experimental performance was not as good as those reported by other studies [11], Most mismatched points occur on sidewalks of narrow roads due to GPS accuracy limitation. Moreover, wheelchairs are hardly equipped with more than one sensor, which makes it impossible to create additional fuzzy logic rules to help matching. With regard to the time performance, this fuzzy logic map-matching algorithm performs reasonably well to meet the demand of real-time navigation, based on the average computation time.

Navigation skills based profiling for collaborative wheelchair control

Peinado et al. [7] present a new approach to proactive collaborative wheelchair control. The system is based on estimating how much help the person needs at each situation and providing just the correct amount. This is achieved by combining robot and hu-man control commands in a reactive fashion after weighting them by their respective local efficiency. Consequently, the better the person drives, the more control he/she is awarded with. In order to predict how much help users may need in advance rather than waiting for them to decrease in efficiency, their skills to deal with each situation are estimated with respect to a baseline driver profile to increase assistance when needed. Situations are characterized at reactive level to keep a reduced set. At reactive level, any situation has to be characterized only within the range of the agent’s sensors. For navigation,complex scenario can be reduced to a set of relatively simple situations that is complete, i.e. fully describes all possible obstacle configurations, and robot and goal locations; and is exclusive. Using a binary decision tree they obtain the regions high-lighted in (Figure 1.3): High Safety or Low Safety Goal in Region (HSGR and LSGR) if the goal is in line of view, High Safety Wide or Narrow Region (HSWR and HSNR), if obstacles around are not too close and Low Safety 1 side and 2 sides (LS1 and LS2), if obstacles are close on one or both sides of the mobile. These situations are represented in Figure (Figure 1.3), where the area considered as Low Safety area around the robot is printed in yellow, the area free of obstacles of the robot is printed in green, obstacles are printed in black and the goal to reach in red.
Assuming that any location of the environment belongs to one of these classes, a user can be profiled by estimating how well he/she solves each of these situations. If someone is able to solve all 6 situations with a reasonable efficiency, it will validate the efficiency to deal with any complex environment and does not require much help. As the word ’rea-sonable’ still not defined yet, a series of experiments took place by gathering driving data from around 80 people, including more than 70 persons with physical and/or cognitive disabilities, and classified the situations they faced while driving into the commented 6 classes. In order to check the proposed model, subjects were told to navigate using their wheelchairs in a standard compliant home for people with disabilities. Results show that collaborative control fills in for skills that users may not have and increases the efficiency of residual ones. They also prove that prediction based modulation improves performance in remaining skills. Future work will focus on correlating low level profile data with high level reasoning techniques via a hybrid control architecture.

READ  Convolutional Neural Network (CNN)

Table of contents :

Declaration of Authorship
1 BEWHEELI project : Why, where and when 
1.1 Introduction
1.2 Related works
1.2.1 Haptic Obstacle Avoidance for Intuitive Powered Wheelchair Navigation
1.2.2 GPS and fuzzy logic map matching for wheelchair navigation
1.2.3 Navigation skills based profiling for collaborative wheelchair control
1.2.4 Discussion
1.3 Emotion, mental fatigue and EEG technology
1.3.1 EEG technology and Brain-Computer interfaces
1.3.2 Mental fatigue, P300 and SSVEP
1.3.3 Emotions
1.4 Gaze/brain based wheelchair command
1.4.1 Gaze based wheelchair command
Wheelesly project
Chern project
Bartolein project
1.4.2 Brain based wheelchair command
BCW project
Toyota project
Muller project
1.5 The BEWHEELI project setup
1.5.1 Materials used for environmental setup
Hardware framework
Virtual world
1.5.2 System framework
1.6 Conclusion
2 Emotion integration in the service of wheelchair control 
2.1 Introduction
2.2 System framework
2.3 Experimental setup for emotion detection
2.3.1 Participants
2.3.2 Procedure
2.3.3 Subjective rating analysis
2.3.4 Feature extraction
Welch method
Discrete wavelets transform
2.3.5 Feature selection
Principal component analysis
Genetic algorithm
Linear Discriminant Analysis (LDA)
Multi layer perceptron (MLP)
Support vector machine (SVM)
2.3.6 Results
2.4 Implementation
2.4.1 Experimental setup
Hardware framework
Virtual world
2.4.2 Procedure
2.4.3 Results
Obstacles hit
Navigation path
Points of gaze
2.4.4 Discussion
2.5 Conclusions
3 Influence of fatigue on wheelchair navigation 
3.1 Introduction
3.2 experimental setup for P300 and SSVEP
3.2.1 Experimental setup
3.2.2 procedure
3.2.3 Classification
3.3 Results
3.3.1 Correlation between fatigue and environmental parameters
3.3.2 Correlation between ERP parameters and subjective ratings
3.3.3 Fusion of data using evidential theory
3.4 Navigation mode switching decision
3.5 conclusion
A A proof of concept: Comparison between gaze based and gaze-brain based navigations 
A.1 Introduction
A.2 BEWHEELI interface description
A.3 Experiment description
A.4 Results and discussion
B Feature extraction and selection techniques : an overview 
B.1 Introduction
B.2 Spectral estimation methods
B.2.1 The periodogram
B.2.2 The modified periodogram
B.2.3 Bartlett Method
B.2.4 Welch method
B.2.5 Discrete wavelets transform
B.3 Selection features methods
B.3.1 Principal Component Analysis (PCA)
B.3.2 Genetic Algorithm (GA)
B.4 Conclusion
C Information fusion theories: State of the art 
C.1 Introduction
C.2 Data fusion challenges
C.3 Data fusion algorithms
C.3.1 Probabilistic fusion
C.3.2 Evidential belief reasoning
C.3.3 Fusion and fuzzy reasoning
Overview of fuzzy logic concept
Fuzzy set
Fuzzy variable
Fuzzy operators
Membership Degree and Probability
Fuzzy logic process
Fuzzy rules
Rules Inference Engine
Fuzzy logic multi-modal data fusion approach
C.3.4 Possibilistic fusion
C.4 Comparison of different theories
C.5 conclusion


Related Posts