AI and ML for Robotics
AI in Robotics
Lecture -3
Dr. Ashok Tadamalle
P rofessor
Department of Mechanical Engineering
DJSCE, Vile Parle, Mumbai
1
Machine Learning in Robotics : Module II
Supervised learning, Unsupervised learning, Reinforcement learning, Deep learning,
Automated Machine Learning, Convergence of IoT & ML, ML algorithms, Classification,
Clustering, Prediction, Motion Heuristics, Types, State space search techniques, Graph
theory techniques, AND/OR graphs, Breadth first search techniques, Hill Climbing, Best
first search techniques, Semantic networks and petri-nets, Dijkstra’s algorithm, Wide
Path Motion Heuristics Method of Path Planning, Sophisticated Motion Heuristics, A
case study to solve a typical robotic problem using ML, Problems.
Automated Machine Learning
(AutoML)
Introduction
What is AutoML?
• Automated Machine Learning (AutoML) automates the end-to-end process of
applying machine learning to real-world problems.
• Reduces the need for expert
data scientists.
• Improves efficiency and
accuracy of models.
• Automates data
preprocessing, feature
engineering, model selection,
and hyper parameter tuning.
• Challenges in Traditional Machine
Learning
• Requires domain expertise.
• Time-consuming model
selection and tuning.
• High computational costs.
• How AutoML Solves These
Challenges
• Automates pipeline
construction.
• Optimizes performance without
manual intervention.
• Democratizes AI by making it
accessible to non-experts.
AutoML Workflow
• Data Preprocessing
• Automated data cleaning, missing value handling, outlier detection.
• Feature Engineering
• Automatic feature selection and transformation.
• Model Selection
• Tries multiple models (Decision Trees, Random Forest, Neural Networks, etc.).
• Hyperparameter Tuning
• Optimizes parameters for best model performance.
• Model Evaluation & Deployment
• Evaluates and selects the best model, then deploys it
Popular AutoML Tools
• Google AutoML
• H2O.ai
• TPOT (Tree-based Pipeline Optimization Tool)
• Auto-sklearn
• Microsoft Azure AutoML
How Google AutoML Works?
• Upload Data: Structured/unstructured data input.
• Model Training: Uses Google’s neural architecture
search.
• Evaluation: Generates accuracy reports.
• Deployment: Model is deployed as an API
TPOT - Automated Machine Learning in Python
• How TPOT Works
• Uses genetic programming to
optimize ML models.
• Advantages
• Finds best pipeline automatically.
• Improves accuracy and reduces
manual effort.
Image Reference:
TPOT workflow diagram
AutoML Use Cases
Industry-wise applications of AutoML
Manufacturing :Predictive maintenance. Healthcare to Predicting diseases with patient data.
Finance: for Fraud detection and risk analysis. E-commerce -Product recommendations
Limitations of AutoML
• Lack of Explainability
• Difficult to interpret the
models.
• Computationally Expensive
• Requires high processing
power.
• Not Always the Best Model
• Sometimes manual fine-
tuning is still required.
Future of AutoML
• Integration with Edge AI
• Better Interpretability & Explainability
• More Efficient Model Search Strategies
• AI-Generated Code for ML Models
Conclusion
• AutoML simplifies machine learning and makes it accessible to all.
• It automates the entire ML pipeline from data preprocessing to model deployment.
• While powerful, AutoML still has challenges like computational cost and
explainability.
Introduction to Machine Learning
• What is Machine Learning
(ML)?
A subset of AI that enables
systems to learn and improve
from experience without explicit
programming.
• Why is ML Important?
• Automates decision-making.
• Improves efficiency in
various domains.
• Types of ML Algorithms
• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning
Categories of Machine Learning Algorithms
• Supervised Learning
• Learns from labeled data.
• Unsupervised Learning
• Finds hidden patterns in data.
• Reinforcement Learning
• Learns through rewards and
punishments.
