Configuration Space, SLAM and Object Recognition

This article explains the concepts of Configuration Space, SLAM, and Object Recognition in robotics. It describes how Configuration Space helps in motion planning, how SLAM is used for mapping, and how Object Recognition is applied for detecting and classifying objects. The article also discusses the challenges and applications of these concepts in different environments.

Running head: CONFIGURATION SPACE, SLAM, AND OBJECT RECOGNITION
1
Defining Configuration Space, SLAM and Object recognition
Student name
Unit title
Institution

CONFIGURATION SPACE, SLAM AND OBJECT RECOGNITION 2
Configuration Space (Cs) in robotics is the space of probable positions a robot could
attain. Also, identified as state space, it can be described as the set of possible transformations
applicable to a robot. Clearly understanding the configuration space assists in solving numerous
motion planning challenges that seem different in terms of kinematics and geometry. Most
concepts involving Cs are borrowed directly from mathematics, especially from topology
(Taylor, 2019). For a robot to attain the right position within its state space, its parameters, which
are the generalized coordinates defining the configuration of a system, need to satisfy
mathematical constraints. Classical mechanics literature considers the configuration of a system
to be the positions held by all components and are subject to kinematical constraints. A
significant concept under configuration space is obstacles. Motion Planning algorithms strive to
find a path with minimum collision constraints in the Cs for the robot during its transformations.
In certain instances, the configurations defining the parameters of the robotic movements could
result in the robot colliding with obstacles or lead to certain links of the robot to collide with
each other. The manifold of robot transformations where the collision constraints are present is
identified as the obstacle region. (Lavalle, 2006) The remaining space without any collision
constraint is the precise path that the robot should follow to attain its configuration goal from
start to finish.
A robotic arm is an example where configuration space is applied. A robotic arm is made
up of a number of rigid linkages. The configuration space comprises of the position of each
linkage, subject to the constraints of how linkages attach together, and, its range of motion
beyond its obstacle region.
Simultaneous localization and mapping (SLAM) is the process of developing a map using
a robot or unmanned vehicle navigating an environment while using the map it creates. The

CONFIGURATION SPACE, SLAM AND OBJECT RECOGNITION 3
SLAM technique is behind robot mapping. The robot or vehicle plots a course in a region but
still has to identify its position at the same time. The SLAM process applies a complex array of
algorithms and computations to traverse within a previously unknown territory. Sensing is one of
the important factors in developing algorithms in SLAM. A range of sensors exists that lead to
creating different SLAM algorithms which are appropriate to the sensors. Laser scanners are
among these sensors, and they provide details of several points within a region. Sensor models
are broadly divided into two approaches, the raw-data and landmark-based approaches
(MAXWEL, 2013 ). Landmarks are the unique identifiable object in a region whose position can
easily be picked by a sensor such as radio beacons. The raw-data approach does not assume that
any landmark can be identified. Instead, it directly models configurations as functions of the
location. Optical sensors could be one-dimensional, 2D, 3D High definition LiDAR, 3D Flash
LIDAR as well as 3D sonar sensors. Currently, intense research is being conducted into visual
SLAM using primarily visual (camera) sensors. This is due to the rising ubiquity of cameras like
the ones in mobile gadgets. Latest forms of SLAM include tactile SLAM which senses by local
touch only and radar SLAM.
A practicable area where SLAM is applied is in underwater environments and in the
airspace. Robots use SLAM in such environments to assist researchers in learning more about
the region by mapping out the landmarks (Thorpe, 2019). Moving objects are, however, a major
challenge of SLAM since they create changes in the environment being surveyed, resulting in
uncertainties in mapping.
Object recognition in robotics involves artificial intelligence that detects and classifies all
occurrences of an object type. An appropriate object recognition technique should detect an
object even in variations of scale, orientation, position, environment, and partial occlusion

CONFIGURATION SPACE, SLAM AND OBJECT RECOGNITION 4
(Hardesty, 2015). The methods applied by robotics equipment in object detection are classified
into how they recognize objects (Machine vision performance) and the amount of time they
require to recognize an image (efficiency). Machine Vision performance works by eliminating
the image parts that fail to match various predefined objects. The efficiency method in object
recognition works by checking the presence or absence of a distinctive class in an image. Thus, if
the image environment is recognizable, the expected object is assigned higher priorities and
detection efficiency. A major challenge in object recognition for robots is consuming a lot of
time in searching through its database after it learns to recognize a large number of items
(Martínez, 2017). The robot's object categories accumulate each instance progressively it
recognizes a new object making it time-consuming to identify the same object while conducting
a search the next time. Object recognition machine intelligence is being applied in the security
field for face recognition which assists in identifying suspects.
All these three concepts, Configuration space, SLAM, and object recognition, are
correlated in that different applications apply these concepts to function as predicted
successfully. Object recognition is a fundamental area applied in SLAM while SLAM is used to
successfully develop the coordinates needed to create the parameters in which a robot can
navigate without collisions in unknown environments.

CONFIGURATION SPACE, SLAM AND OBJECT RECOGNITION 5
References
Hardesty, L. (2015, July 23). Object recognition for robots. Retrieved from MIT News Office:
http://news.mit.edu/2015/object-recognition-robots-0724
Lavalle, S. (2006). Configuration Space Obstacles. Retrieved from Cambridge University Press:
http://planning.cs.uiuc.edu/node156.html
Martínez, E. a. (2017). Object Detection and Recognition for Assistive Robots. IEEE Robotics &
Automation Magazine, 45-99.
MAXWEL, R. (2013, JANUARY, 15). Robotic Mapping: Simultaneous Localization and
Mapping (SLAM). Retrieved from GISlounge: https://www.gislounge.com/robotic-
mapping-simultaneous-localization-and-mapping-slam/
Taylor, C. (2019). Introduction to Configuration Space. Retrieved from Coursera Inc:
https://www.coursera.org/lecture/robotics-motion-planning/2-1-introduction-to-
configuration-space-0auId
Thorpe, C.-C. W. (2019). Simultaneous Localization and Mapping with Detection and Tracking
of Moving Objects. Pittsburgh,: Robotics Institute, Carnegie Mellon University, .