The Mecanum wheel was designed in Sweden in 1975 [1]. Using four of these
wheels provides omni-directional movement for a vehicle without needing a conventional
steering system [2–6]. The wheel itself consists of a hub carrying a number
of free moving rollers angled at 45 about the hub’s circumference. The rollers are
shaped such that the overall side profile of the wheel is circular. However, wheel slip
is a common problem with the Mecanum wheel, particularly when the robot moves
sidewise, as it has only one roller with a single point of ground contact at any one
time. This severe slippage prevents the most popular dead-reckoning method, using
rotary shaft encoders [7,8] from being performed well on the Mecanum robot. To
cope with the problem, visual dead-reckoning was used as a slip-resilient sensor [9–
11]. This technique, also used in optical mice, makes use of an on-board videocamera
continuously capturing frames of the ground beneath and image processing
hardware on the robot determining the speed and direction in which the current
frame has moved relative to the previous frame thus allowing the speed and direction
of that point of reference to be calculated. However, visual dead-reckoning using a
single camera or optical mouse can not provide all three-degree-of-freedom positional
information for robot navigation and motion control. Fixed line following is
the simplest and most reliable solution, yet is also the most limiting. A physical line is
marked on the ground along the path which the robot is to follow [7,8]. For a robot
that is set up in a fixed location for a set task this system is effective but for a research
robot with omni-directional capability this approach is seen to be a primitive, though
still viable, option.
This paper presents a final-year project recently completed at Massey University,
New Zealand. The research started upon an existing omni-directional platform built
of a box-like aluminium chassis, four electric window winder motors and four
Mecanum wheels [12]. The aim of this project was to provide the platform with
motion control that could be programmed to accommodate various robotic
behaviours specified. With respect to the path following behaviour, two optical mice
were attached to give positional feedback for closed-loop control and dead-reckoning
for navigation and a Mitsubishi M16C/62 microcontroller was interfaced and
programmed to implement robotic behaviours. A closed-loop control in Cartesian
space was proposed to control x- and y-movement and rotation motions of the
robot.
As this was a project incorporating mechanical, electrical and software development,
a mechatronics design principle was applied. The different areas were developed
synergistically thus allowing interactions between the disciplines to be viewed
and managed. It also meant that all three core disciplines needed to be developed to a
certain stage before any one area could be further worked on. Although it was
physically possible to use other means to develop the core areas independently, a
synergistic approach tends to be more efficient. Even though this parallel design
approach was used, the areas of development shall be discussed in sections assuming
that other sections have already been completed to a certain level and are referenced
where necessary.