Combining Local and Global Optimisation
for Virtual Camera Control
Paolo Burelli, Student Member, IEEE, and Georgios N. Yannakakis, Member, IEEE
Abstract— Controlling a virtual camera in 3D computer
games is a complex task. The camera is required to react to
dynamically changing environments and produce high quality
visual results and smooth animations. This paper proposes an
approach that combines local and global search to solve the
virtual camera control problem. The automatic camera control
problem is described and it is decomposed into sub-problems;
then a hierarchical architecture that solves each sub-problem
using the most appropriate optimisation technique is proposed.
The approach is compared to pure local search solutions to
showcase the advantages of the proposed architecture in terms
of visual performance and robustness.
I. INTRODUCTION
Camera control has a deep impact on player experience
and enjoyability in games [24]. The camera represents the
point-of-view of the player through which she perceives the
game world and gets feedback on her actions.
Camera settings for games are usually directly controlled
by the player or statically predefined by designers. Direct
control of the camera by the player increases the complex-
ity of the gameplay interaction and reduces the designer’s
control on game storytelling (e.g. the player might point the
camera towards an area which reveals unwanted informa-
tion). On the other hand, a designer-driven camera control
releases the player from the burden of controlling the point
of view, but often generates undesired camera behaviours
(e.g. the player is hidden behind an object). Moreover, if the
content of the game is procedurally generated, the designer
might not have any information to define a priori the camera
positions and movements.
Automatic camera control aims to define an abstraction
layer that permits the designers to instruct the camera with
high-level and environment-independent rules. The camera
controller should dynamically and efficiently translate these
rules into camera movements while the player plays the
game.
Several techniques for automatic camera control have been
proposed in the past — the reader is referred to [12] for a
comprehensive review. The most common approaches model
the camera control problem as a constraint satisfaction or
optimisation problem. These approaches allow the designer
to define a set of requirements on the frames that the camera
should produce and on the camera motion. Depending on
the approach, the controller positions and animates one or
more virtual cameras that attempt to satisfy the predefined
requirements.
Authors are with the Center for Computer Games Research, IT University
of Copenhagen, Rued Langgaards Vej 7, DK-2300 Copenhagen S, Denmark.
Emails: {pabu, yannakakis}@itu.dk
Finding the best camera positions and movements that
satisfy the designer’s requirements in a dynamic three di-
mensional environment is a complex task. Evaluation of
frame properties, such as object visibility, is computationally
expensive and the evaluation functions often generate terrains
that are very rough for a search algorithm to explore [10].
Moreover, the camera is required to react in real-time at
dynamic change of the environment or player action which
forces the computation time to be constrained within small
time intervals (e.g. 16.7 ms for 60 fps).
Pure global optimisation approaches [23], [11], [8], [17]
are capable of producing well composed shots with respect
to designer requirements. However, their high computational
cost makes them inappropriate for real-time interactive ap-
plications such as games.
Local search approaches [4], [6], [9], on the other hand,
offer real-time performance and allow the designer to control
also camera motion parameters such as speed and accelera-
tion but they tend to stick to local optimum solutions.
This paper proposes an approach to the automatic camera
control problem that employs local search, global search
and path planning to generate smooth camera animations
and well composed shots. The camera control problem is
decomposed into smaller tasks and different techniques are
used to perform different tasks overcoming the limitations of
the pure approaches.
The approach proposed extends and draws upon one of
the author’s earlier work on local search based camera
control [9] and on visibility optimisation [10]. The new
camera controller is built upon a layer of three components:
the local search algorithm proposed in [9] is used to find
the best camera configuration; a stochastic population-based
global search algorithm is employed to avoid premature
convergences to local minima and a real-time efficient path
planning algorithm is designed to generate smooth camera
animations and to control camera movements.
The CamOn camera control system, that embeds all three
modules, is evaluated through three case studies varying in
complexity. Results show that the approach proposed demon-
strates robustness and high visual performance across all the
three case studies. It is also apparent that the computational
cost of this combined approach does not differ significantly
from pure local search, but the visual performance, both in
terms of composition and animation, is significantly higher.
This paper is innovative in that it introduces an efficient
and reliable hybrid solution to the automatic camera control
problem coupling local with global search. The proposed ap-
proach successfully combines different optimisation methods