Anti-Gravity Movement
Anti-Gravity movement (or in the field of robotics it is also known as potential field motion planning) is a path finding technique, the idea is to use the formula for gravity to calculate a path, unlike graph search algorithms where we calculate the full path between A and B, anti-gravity movement calculates the path one step at a time. This makes it ideal for dynamic worlds because there is no risk that a path becomes obsolete due to changes in the world.
The theory behind anti-gravity is actually quite simple, we place imaginary points on the map called gravity points that we want to be repelled or attracted to, each gravity point has a force associated with it which determines how much we want to be attracted or repelled from the point allowing us to easily create movement patterns. At each step, for each point we calculate gravitational force acting on us from that point using the formula \(\frac{strength}{distance\^n}\), summing all the forces from all the points on the map will yield the total gravitational force acting on us then we let it push us in whatever direction it push us in.
All vector operations used are from my vector-2d library you can grab it from clojars.
(ns gravity.core
(:refer-clojure :exclude [+ - * =])
(:use (clojure.contrib.generic [arithmetic :only [+ - *]]
[comparison :only [=]]))
(:use [vector-2d.core]))
(defn gravity-vector [u v]
(let [force (/ 2000 (Math/pow (magnitude (- u v)) 2))
angle (bearing u v)]
(vector-2d (* (Math/sin angle) force) (* (Math/cos angle) force))))
(defn total-gravitational-force [state]
(apply + (map #(gravity-vector (:player @state) %) (:obstacles @state))))
Using the formula \(\frac{strength}{distance\^n}\), in this case strength is 2000 and n is 2, we calculate the force then using the angle between v and u we create a new vector from v to u which pushes us away from v. Experimenting with different strength and n values will yield different behaviours. i.e increasing n will cause you to avoid points when you get very close to them. Doing this for all the points on the map and then summing them up yields total gravitational force.
(defn seek [state]
(normalize (+ (normalize (- (:target @state) (:player @state)))
(total-gravitational-force state))))
(defn steer [state]
(when (> (dist (:player @state) (:target @state)) 1)
(dosync (alter state assoc :player (+ (:player @state) (seek state))))))
To actually go where we want to go, we subtract our current position from the target, normalizing this resulting vector gives us a unit vector pointing to where we want to go, then we add total gravitational force to the direction vector which will push us away from the obstacles and towards our target.
(defn circle [g pt rad color]
(let [[x y] (vals pt)
offset (int (/ rad 2))
x (- x offset)
y (- y offset)]
(doto g
(.setColor color)
(.fill (java.awt.geom.Ellipse2D$Double. x y rad rad)))))
(defn board [state]
(proxy [javax.swing.JPanel
java.awt.event.ActionListener] []
(paintComponent
[g]
(.setColor g java.awt.Color/WHITE)
(.fillRect g 0 0 (.getWidth this) (.getHeight this))
(circle g (:player @state) 20 java.awt.Color/GREEN)
(doseq [pt (:obstacles @state)]
(circle g pt 60 java.awt.Color/BLUE)))
(actionPerformed
[e]
(steer state)
(.repaint this))))
(defn mouse-listener [state]
(let [dragging (ref nil)]
(proxy [java.awt.event.MouseAdapter
java.awt.event.MouseMotionListener] []
(mousePressed
[e]
(dosync (alter state assoc :target
(vector-2d (.getX e) (.getY e))))))))
(defn frame []
(let [state (ref {:player (vector-2d 325 200) :target (vector-2d 325 200)
:obstacles [(vector-2d 379 90)
(vector-2d 135 223) (vector-2d 560 290)
(vector-2d 550 120) (vector-2d 145 70)
(vector-2d 350 320) (vector-2d 60 50)
(vector-2d 210 313) (vector-2d 450 210)]})
board (board state)
timer (javax.swing.Timer. 5 board)
mouse-listener (mouse-listener state)]
(.addMouseListener board mouse-listener)
(.start timer)
(doto (javax.swing.JFrame.)
(.setAlwaysOnTop true)
(.add board)
(.setSize 650 400)
(.setVisible true))))