COMP0037: Robotic Systems
A Very Brief Introduction to Robotics
Simon Julier
Email:
s.julier@ucl.ac.uk
Office hours:
15:00-17:00, Thursdays, 66-72 GS 5th Floor
4 Factors of Rational Agents
• The performance measures that define the criteria
of success
• The agent’s prior knowledge of the environment
• The actions the agent can perform
• The agent’s percept sequence to date
2
4 Factors of Rational Agents
Percepts
Sensors
Environment
Agent
Actuators
Actions
3
Rational Agent: Trading System Support
5
Rational Agent: Augmented Reality System
6
Rational Agents: Robots
https://www.youtube.com/watch?v=BGOUSvaQcBs 7
Scope of Motion Planning in this Module
• We can’t investigate all of the planning and
actions which really complicated robots undertake
– The KAIST robot has more than 15 degrees of freedom
• In this module, we will consider the focused
problem of motion planning of warehouse robots
8
Motion Planning
• … A fundamental need in robotics is to have algorithms
that convert high-level specifications of tasks from humans
into low-level descriptions of how to move. The terms
motion planning and trajectory planning are often used for
these kinds of problems.
- From Chapter 1 of “Planning Algorithms”, LeValle, 2006
9
Warehouse Robotics: Automatic Forklifts
Estatroll Paqueto 10
Warehouse Robotics: Self-Moving Shelves
Kiva Systems (Amazon) 11
Warehouse Robotics: Modular Storage
http://autostoresystem.com/ 12
Warehouse Robotics: Tote Carrying
Locus Robotics, https://www.youtube.com/watch?v=pjfYr_po6fE
13
Warehouse Robotics: Robot Tasks
• The robots are tasked with travelling between a
start and end goal in a warehouse
• The warehouse contains both fixed stationary
objects and mobile objects
• Assumptions:
– The robot always knows its location
– The robot perceives its nearby environment
– The robot controls its movement by specifying the
current on the motors of each of its wheels
14
Robot Localisation
From
https://sites.google.com/a/webmail.korea.ac.kr/intelligent-robot-laboratory/navigatio15
n/mono-slam
Environmental Scanning
16
Robot Motion
Kinematics of a differential drive robot, from
http://rossum.sourceforge.net/papers/DiffSteer/ 17
Motion Planner Challenges
• The planner has incomplete information about:
– The environment and whether there are blockages
– The model which describes how the robot moves
• Therefore, the planner must:
– Be reactive and revise its plans as necessary
– Operate in real-time
• Hierarchical planners often used
18
Hierarchical Motion Planner
From http://dnc.tamu.edu/projects/cooperative/images/ccma.png 19
Hierarchical Motion Planner
World High Level
Model Tasking
Global
Planner
Local
Perception
Planner
Motion Inner Loop
Controller Vehicle Control
Actual
World Vehicle
Modified from “Reactive
Motion Planning for a Car”, 20
M. Hasson, C. Laugier
Global Planner
• Periodically generates a top-level plan which is
often a series of tasks
• The tasks often consists of a geometric trajectory
– Waypoints
– Time constraints
• It is formulated using current knowledge about the
world and high level objectives
• Tries to be optimal, but can be expensive and
slow to run
• Often decomposes the problem into small parts
21
Hierarchical Task Network Global Planner
• One approach is to use a Hierarchical Task
Network (HTN)
• The problem is to decompose high level tasks into
a set of primitive tasks that the local planner can
execute on its own
• The approach is to model everything as a task
network and decompose it recursively
22
Hierarchical Task Network Global Planner
Task 1 Task 2 Task 3
23
Hierarchical Task Network Global Planner
Task 2.1 Task 2 Task 2.2
24
Hierarchical Task Network Global Planner
Task 1 Task 2.1
Task 2.2 Task 3
25
Hierarchical Task Network Global Planner
Task 1 Task 2.1.1 Task 2.1.2
Task 2.2 Task 3
26
Local Planner
• Is given the set of primitive tasks to execute
• Continually monitors the environment and creates
short term plans
• These plans are often smoothed trajectories which
provide optimal reference signals for the
controllers
• Can handle “small” changes to the environment
– Prevents collisions, plan around obstacles
• Often not optimal
27
Local Planning and Obstacle Avoidance
From http://pubs.sciepub.com/automation/1/1/2/figure/17
28
Local Planning and Obstacle Avoidance
From http://pubs.sciepub.com/automation/1/1/2/figure/17
29
Inner Loop Vehicle Control
• Is given a set of primitive actions or paths to
complete
• Very tight control loops at the level of low-level
motor control
• Uses detailed models of robot physics and
dynamics to work out the control sequences used
• Only reactive and responds to things like collision
events and hardware failures
30
Summary
• Robots are a kind of rational agent
• Motion planning is key in many robot operations
• In general, motion planning can be very
complicated and high dimensional
• Reactive, hierarchical planning systems have to
be used
• We will be looking at simple warehouse robotics
• The first step will be to formulate the path planning
problem
31
