01_Introduction

• intelligence - collective term in psychology for human cognitive performance

• -> no agreement on how to distinguish betweeen them

• => intelligence requires separation from different categories

• emotional intelligence

  • feelings, empathy, motivation

• creative itnelligence

  • imagination, innovation, intuition

• methodical intelligence

  • structure, system, discipline

• analytical intelligence

  • critical analysis, logic, strategic thinking

develop machines that behave as if they had intelligence

• Thinking rationally

  • laws of thought, logic, correct reasoning

• thinking humanly

  • thought proceduer, human performance

• acting rationally

  • acting agents, act autonomously, persost long, adapt

• acting humanly

  • turing test, natural language, knowledge storage, ML

1. Reasoning and Problem solvning

2. Knowledge representation

3. Planning

4. Learning

5. NLP

6. Perception

7. Motion and manipulation

8. social intelligence

9. general intelligence

• ability to setp-by-setp reason by logical dedutctions with uncertainity

• e.g. solve math equations… using formal logic and setp-by-step deduction…

• searching (tree, dijkstra, kruskal, nearest neighbor, A*)

• optimization (linear programming, quadratic programming, heuristic, bayesian optimization)

• evolutionary computation (genetic algo, particle swarm, ant colony)

represent information about the world for e.g. an intelligent agent

• implement knowledge as axioms/sentences with facts and rules about the world

• locig (propositional logic, first order logic, fuzzy logic)

• how to deal with uncertainigy when deciding on actions to take? (represent world and future for an agent)

• searching ( tree, dijkstra, …) -> search through many possible solutions…

• agent-systems (multi agent, intelligent agent)

uncertainity reasoning (bayesian, hidden markov, kalman) -> operate with incomplete information

• allow for a computer to find patterns in data

• learn from data and not follow static programming…

• supervised, unsupervised

• allow for a computer to understand human language

• -> e.g. understand hand-written sources,

• -> interaction between human and computer… (e.g. siri)

• logic

• classical machine learning

• deep learning

• make sense of data an agent gets from the environment

• -> paint picture of environent based on this data…

• classical computer vision

• moden computer vision

• machine perception

• plan behavior and motion of agent

• choose locomotion (rolling, walking)

• sense the environment

• conntrol actuators

• behavioral planning (what should i do) - logic-based (state machine), knowledge based (network graph)

• motion planning (how can i acheive sth) - search algo, optimization algo

• control (steering and conntrol of all actuators) - PID, model predictive congtrol

• understand and reprodice social skills

  • self awareness, self managemet, social awareness, soclal management

• recognize, interpret, process and simulate human effects

• -> speech detection, facial affect detection, body gesture detection, physical monitoring…

• database - logic based (state machine), knowledge based

• classification - what emotion could this be?

  • SVM, k-NN, DL

• game theory - mathematical interaction etween intelligent rational decision makers

  • cooperative game, simultaneous game, evolutionary gmae

• computatinal complexity

  • lots of problems NP hard

• information complexity

  • limited informaiton

  • uncertainty

  • knowledge acquisition required

• data, labeled data, knowledge available

  • Big Data

• new AI algos available

  • Deep Learning

• computer power available

  • GPU

• big data analysis

• faster matrix muliplication

• machine translaiton

• NLP

• securiry (face regognition)

• …

• automotive development

• vehicle functions (e.g. predictive maintenance)

• complete vehicles (autonomous driving)

• automotive production

• safety improvement (most accidents due to human error)

• comfort improvement

• energy saving (better planned velocity and trajectories)

• traffic reduction (exchange VXV for traffic adaption)

• new mobility services (google transport, ai taxi)

• new software function development

• level 0

  • no autonomous features

• level 1

  • cars can handle one task at a time (e.g. autonomous breaking)

• level 2

  • at least two automated functinos)

• level 3

  • car can handle dynamic driving tasks but mithg still reqhire interventions (e.g. spurhalteassistent, … tempostat und auffahrbremsen…)

• level 4

  • officially driverless in certain environments

• level 5

  • operate entirely on their own without any driver presence

• navigation - hours to minutes

• path following - minutes to seconds

• control / stabilization (e.g. object avoidance) seconds to miliseconds

• world is complex and dynamic place (weather, cars, pedestrians, animals,…)

• => use ML algos to teach cars how to drive like a human being

• sensing (taking infor from environmetn)

• perception (making sense of the information to paint a picture of the environment)

• planning (use the gained image of the environment to plan the behavior of the car and e.g. make predictions of the future)

• control (act upon the planning decisions)

• radar

• lidar

• ultrasonic

• camera

• gps

• IMU (inertia -> beschleuningung…)

  
  

• Detection

  • lane detection

  • traffic light detection & classification

  • traffic sign detection & classification

  • object detection & tracking

  • free space detection

• Localization

• route planning

• prediction

• behavior planning

• trajectory planning

• PID

• MPC

• Others

• Camera

• Radar

• Lidar

• GPS

• Othres