Simulating object-centred neglect with head-centred coding of space based on non-linear gaze-dependent units