目的

　　此例子的AI效果就是要车子在一条连续的封闭的曲线上面不停的自己循环移动。

其它

　　对车子运动控制，请看上一篇http://www.cnblogs.com/zkzk945/p/5146474.html

基本思路

　　利用样条曲线(catmull-rom spline)将路径点通过插值形成连续的曲线，根据车子离开路径起始点的距离计算出车子在路径上的目标点，目标点决定了车子的朝向、油门、刹车。

target.position = circuit.GetRoutePoint(progressDistance + m_lookAheadForTargetOffset +                                                m_lookAheadForTargetFactor*speed).pos;        target.rotation = Quaternion.LookRotation(circuit.GetRoutePoint(progressDistance + m_lookAheadForTargetOffset +                                                m_lookAheadForTargetFactor * speed).dir);

以上代码就是根据和初始点的距离计算目标点，包括了目标点的位置和朝向。

progressPoint = circuit.GetRoutePoint(progressDistance);        //Debug.Log(progressDistance);        Vector3 progressDelta = progressPoint.pos - transform.position;        if (Vector3.Dot(progressDelta, progressPoint.dir) < 0)        {            progressDistance += progressDelta.magnitude*0.5f;        }

以上代码是为了判断车子是否在目标点的前面，如果是，那么增加progressDistance，可以让target目标点定位到车子的前方去，这样target点就会在样条曲线上面循环移动，那么车子也会尾随其移动。

//遇到转角或者控制节点，减速处理                float desiredSpeed = m_carController.MaxSpeed;                Vector3 delta = m_target.position - transform.position;                float distanceCautiousFactor = Mathf.InverseLerp(m_cautiousMaxDistance, 0, delta.magnitude);                float spinningAngle = m_rigidbody.angularVelocity.magnitude*m_cautiousMaxDistance;                float cautiousnessRequired = Mathf.Max(Mathf.InverseLerp(0, m_cautiousMaxAngle, spinningAngle),                    distanceCautiousFactor);                desiredSpeed = Mathf.Lerp(m_carController.MaxSpeed, m_carController.MaxSpeed*m_cautiousSpeedFactor,                    cautiousnessRequired);

以上代码是在转角点时对期望速度的处理，根据车子和目标点的距离、车子本身的角速度，求得一个紧张系数，当车子距离目标点越紧或者角速度越大，紧张系数也就越大，这样车子就需要减慢速度，以防止翻车。

Vector3 offsetTargetPos = m_target.position;                offsetTargetPos += m_target.right*                                   (Mathf.PerlinNoise(Time.time*m_lateralWanderSpeed, m_randomPerlin)*2 - 1)*                                   m_lateralWanderDistance;                Vector3 localTarget = transform.InverseTransformPoint(offsetTargetPos);                float targetAngle = Mathf.Atan2(localTarget.x, localTarget.z) * Mathf.Rad2Deg;                float steer = Mathf.Clamp(targetAngle * m_steerSensitivity, -1, 1);                float accelBrakeSensitivity = (desiredSpeed < m_carController.CurrentSpeed)                    ? m_brakeSensitivity                    : m_accelSensitivity;                float accle = Mathf.Clamp((desiredSpeed - m_carController.CurrentSpeed)*accelBrakeSensitivity, -1, 1);                //Debug.Log(accle);                //Debug.Log((1 - m_accelWanderAmount) +                //         (Mathf.PerlinNoise(Time.time*m_accelWanderSpeed, m_randomPerlin)*m_accelWanderAmount));                accle *= (1 - m_accelWanderAmount) +                         (Mathf.PerlinNoise(Time.time*m_accelWanderSpeed, m_randomPerlin)*m_accelWanderAmount);                //Debug.Log(accle);                m_carController.Move(steer, accle, accle, 0f);

首先对目标点进行柏林噪声随机，然后根据目标点算出方向盘的方向。然后根据期望速度和当前速度计算出油门或者刹车输入，最后手动传给控制汽车移动的函数，实现了汽车的自动化运行。

源码

http://git.oschina.net/zkzk945/CarAIDemo521