Chapter 10 - Linear Control of Manipulators
Chapter 10 - Linear Control of Manipulators
Outline
2. PID control
Desired workspace
Real Workspace
trajectory
trajectory
𝑥! 𝑦! 𝑧!
x𝑦𝑧
Inverse
Kinematics Forward
Kinematics
The Objective is to design a controller to ensure a good tracking in joint space and workspace
4
v A key feature of control systems is the feedback loop. It involves taking the output and
feeding it back into the system as input. This loop helps in self-correction and maintaining
the system's stability, ensuring it consistently achieves the set goals even in the presence of
disturbances or changes.
v Controllers aim to minimize tracking errors, counteract disturbances, and filter out noise,
ensuring precise alignment with desired paths or values.
6
PID Controller
7
PID Controllers
l This note examines a particular control structure that has become almost
universally used in industrial control.
– I (Integral)
– D (Derivative)
8
PID Controllers
• Four modes of control commonly used for most
applications are:
– Proportional (P)
PID Controllers
We assume that the PID controller is placed in a closed-loop unity feedback system
as shown in figure below. The variable e(t) denotes the joint space tracking error,
which is sent to the PID controller.
The control signal 𝜏(t) from the controller to the robot is equal to:
𝑑𝑒(𝑡)
𝜏 𝑠 = 𝐾! 𝑒(𝑡) + 𝐾" / 𝑒 𝑡 𝑑𝑡 + 𝐾#
𝑑𝑡
PID Controllers
How do the PID parameters affect system dynamics?
𝜏(𝑠)
The effects of increasing each of the PID controller parameters 𝐾! , 𝐾" , and 𝐾# are
summarized in table below
𝑑𝑒 𝜏 = 𝐾# 𝑒 + 𝐾$ 𝐷 %! 𝑒 + 𝐾& 𝐷 " 𝑒
𝜏 = 𝐾# 𝑒 + 𝐾$ 2 𝑒𝑑𝑡 + 𝐾&
𝑑𝑡
15
v Control Flexibility: Highlight the concept of "tuning flexibility" that comes with FOPID
controllers. The fractional orders λ and μ add two more tuning parameters to the
traditional PID, allowing for a much more nuanced control strategy that can be tailored to
the specific needs of a wide range of systems, from highly oscillatory systems to those
with long time delays.
𝑑𝑒
𝜏 = 𝐾# 𝑒 + 𝐾$ 2 𝑒𝑑𝑡 + 𝐾& 𝜏 = 𝐾# 𝑒 + 𝐾$ 𝐷 %! 𝑒 + 𝐾& 𝐷 " 𝑒
𝑑𝑡
𝑇(𝑠) 𝐾$ 𝑇(𝑠) 𝐾$
𝐶 𝑠 = = 𝐾# + + 𝐾& 𝑠 𝐶 𝑠 = = 𝐾# + ! + 𝐾& 𝑠 "
𝐸(𝑠) 𝑠 𝐸(𝑠) 𝑠
17
where 𝑞,𝑀(⋅),𝐶(⋅),𝐺(⋅) and 𝜏 are a set of generalized coordinates, the inertia matrix, the
Coriolis/Centripetal vector, the gravity vector and the control input, respectively.
v One of many goals of a control engineer is to choose the control input 𝜏 so that the robot
should follow a desired trajectory (i.e. denoted as 𝑞𝑑𝑒s or 𝑞𝑑 for short). The
aforementioned equation can be rewritten as
Ø The error dynamic is always deduced by inserting the control law in the dynamic model
𝑞̈ & − 𝑞̈ + 𝐾# 𝑒 + 𝐾& 𝑒̇ = 0 ↔ 𝒆̈ + 𝑲𝒑 𝒆 + 𝑲𝒅 𝒆̇ = 𝟎
19
+
The characteristic equation of this eroor dynamics is given as follows:
det 𝑆𝐼 − 𝐴 = 𝑠 * + 𝐾& s + 𝐾# = 0
It is clear that the error dynamics satisfies the exponential stability for any positive gains 𝐾#
and 𝐾& , and consequently the error converges to the equilibrium point (i.e., qdes→ q).
20