VIP20
本课程属于：005-Artificial Intelligence and Machine Learning Engineering English Technical Courses · 共 20 门相关课程，组合学习更完整。
查看所属VIP全部课程 →
← 返回所属目录 咨询 / 购买

Reinforcement Learning for Engineering Control Systems

30 Chapters
View Course Samples Online
01
Engineering control systems as sequential decision problems
state, action, reward, and dynamics in continuous-time and discrete-time plants.
sequentialdecision
02
Markov decision processes for control engineers
transition kernels, cost-to-go, Bellman operators, and connections to optimal control.
decisionprocesses
03
Model-free value estimation in control loops
Monte Carlo returns, temporal-difference learning, and eligibility traces for online adaptation.
Model-freeestimation
04
Q-learning and SARSA for setpoint regulation
tabular implementations, exploration schedules, and convergence on benchmark plants.
Q-learningSARSA
05
Policy gradient methods for continuous control
REINFORCE, natural gradients, and variance reduction in actuator command spaces.
gradientcontinuous
06
Actor-critic architectures for tracking control
advantage estimation, compatible features, and real-time implementation constraints.
Actor-criticarchitectures
07
Deep deterministic policy gradient (DDPG) for industrial process control
replay buffers, target networks, and low-level actuator interfacing.
deterministicgradient
08
Twin delayed DDPG (TD3) for robust setpoint tracking
delayed policy updates, target policy smoothing, and clipped double-Q learning.
DDPGTD3
09
Soft actor-critic (SAC) for energy-efficient control
maximum entropy objectives, automatic temperature tuning, and stability margins.
actor-criticSAC
10
Proximal policy optimization (PPO) for safety-constrained regulation
trust regions, clipping, and constraint handling in continuous action spaces.
Proximaloptimization
11
Model-based reinforcement learning for control
learned dynamics models, MPC integration, and uncertainty-aware rollouts.
Model-basedreinforcement
12
Linear quadratic regulator (LQR) and RL
policy iteration, Riccati equations, and data-driven approximations for unknown dynamics.
quadraticregulator
13
System identification and RL
subspace methods, neural ODEs, and joint model-policy learning for engineering plants.
identificationsubspace
14
Reward shaping for engineering objectives
sparse rewards, potential-based shaping, and multi-objective trade-offs in control tasks.
objectivespotential-based
15
Safety and constraint satisfaction in RL control
control barrier functions, shielding, and safe exploration with Lyapunov methods.
constraintsatisfaction
16
Robust reinforcement learning for uncertain plants
domain randomization, adversarial disturbances, and worst-case guarantees.
reinforcementuncertain
17
Hierarchical reinforcement learning for multi-level control
options, feudal networks, and task decomposition in process industries.
Hierarchicalreinforcement
18
Multi-agent reinforcement learning for distributed control
cooperative MARL, centralized training with decentralized execution, and consensus.
Multi-agentreinforcement
19
Sim-to-real transfer for control systems
domain adaptation, system delays, actuator nonlinearities, and hardware-in-the-loop validation.
Sim-to-realtransfer
20
RL for PID and gain scheduling
automatic tuning, adaptive gain adjustment, and comparison with classical tuning rules.
PIDscheduling
21
RL for robotic manipulator control
joint-space tracking, impedance control, and contact-rich manipulation tasks.
manipulatorjoint-space
22
RL for power electronics and motor drives
current control, speed regulation, and torque ripple minimization.
electronicsregulation
23
RL for HVAC and building energy management
thermal comfort, energy cost, and demand response with model-free agents.
HVACbuilding
24
RL for autonomous vehicle control
lateral and longitudinal control, trajectory tracking, and safety-critical decision making.
autonomouslongitudinal
25
RL for chemical process control
reactor temperature control, distillation column operation, and constraint-aware policy learning.
chemicaltemperature
26
Offline reinforcement learning for control
batch-constrained Q-learning, conservative Q-learning, and deployment without online interaction.
reinforcementbatch-constrained
27
Imitation learning and RL for control
behavioral cloning, DAgger, and inverse reinforcement learning for controller design.
Imitationbehavioral
28
Meta-reinforcement learning for adaptive control
fast adaptation to changing plant dynamics and task distributions.
Meta-reinforcementadaptive
29
Evaluation and benchmarking of RL controllers
sample efficiency, stability metrics, reproducibility, and industrial case studies.
Evaluationbenchmarking
30
Deployment and lifecycle management of RL controllers
edge inference, monitoring, retraining, and regulatory compliance in engineering systems.
Deploymentlifecycle